1. Introduction

Integrated optical refractive index (RI) sensors have been extensively investigated for a number of applications and play a prominent role in biochemical analysis for their attractive features such as high sensitivity, miniature dimensions, mechanical stability, and immunity to electromagnetic interference [1]. Additionally they offer the prospect of being incorporated in lab-on-a-chip that is capable of doing point-of-care (POC) measurements at an affordable cost. Among various integrated optical devices used for sensing applications, Mach-Zehnder interferometer (MZI) sensors are easy and simple in design, fabrication and measurement, and able to provide high sensitivity, which make them suitable sensing platform for the development of portable POC device for biomedical applications. A variety of material systems for MZI sensors have been reported including silicon oxide [2–4], siliconoxynitride [5], silicon nitride [6–9], silicon-on-insulator (SOI) [10], porous silicon [11], and polymers [12,13]. MZI sensors have also been demonstrated in different waveguide structures, such as conventional strip and rib waveguides [3,5,7,10,12,13], arrow waveguide [4], plasmonic waveguide [14], and slot waveguide [8,9].

A conventional MZI sensor configuration is based on two channels formed by one splitter and one combiner connected by a pair of separated straight waveguides – the sensing and the reference arm. A relatively wide separation between the sensing and the reference arms is needed in order to avoid evanescent coupling between the waveguides and inevitably requires additional bending or coupling structures for the signals to be split into or recombined from the channels, thus limiting the capability to further reduce the device size. Single-channel MZI sensor has been proposed by using the interference between two polarizations of a single-mode waveguide [6], however additional bulky optical component such as linear polarizer is required. Single-channel MZI sensors have also been investigated and demonstrated in a two-transversal-mode waveguide structure using two modes of the same polarization [15,16]. The utilization of transversal modes needs a discontinuous junction in vertical direction for mode excitation, which requires multiple-step lithography and etching process. Careful etching depth control is also needed for appropriate mode excitation and large interference visibility. In addition, for all single-channel MZI sensors based on the use of two modes of the same waveguide, several centimeter-long sensing waveguide is usually required for high sensitivity since both modes can sense RI change and their accumulated phase changes cancel out. To reduce the size of the sensing area, single-mode spiral waveguide has been used in traditional two-channel MZI configuration, however still requires a balanced spiral arm and additional structures for light splitting and combining [10].

In this paper, we propose and demonstrate a compact single-channel MZI sensor by using a two-lateral-mode spiral sensing waveguide. Two-lateral-mode waveguide offers the advantage of simpler fabrication using single-step lithography and etching process over two-transversal-mode waveguide structure. A simple discontinuity is utilized for the excitation of the two lateral modes and the optimized excited power ratio for high interference visibility is easily controlled by the offset distance of the discontinuity. The two-mode sensing waveguide is folded in a spiral layout to obtain high sensitivity of a long sensing arm while providing a compact sensing area compatible with commercial spotting machine. The paper is organized as follows: in Section 2, we describe the working principle and design of the sensor. In Section 3, the sensor fabrication is introduced and the effect of the discontinuity offset distance on the interference visibility is studied. Then the bulk and surface sensitivity of the sensor are characterized respectively by sodium chloride (NaCl) solutions of different concentrations and polyelectroltye multilayer deposition. In Section 4, the biosensing capability of the sensor is demonstrated and verified by the measurement of biotin–streptavidin interaction of different concentrations. Discussion and concluding remarks are given in Section 5.

2. Device structure and design

Figures 1(a) shows the schematic diagram of the sensor, which consists of two single-mode waveguides connected by a two-lateral-mode spiral sensing waveguide through two discontinuous junctions. Two grating couplers are placed at the input and output ends for light coupling into (and out of) the sensor. Only transverse-electric (TE) polarization is considered in this paper for the sake of simplicity and the transverse-magnetic (TM) polarization works similarly. Light is first launched into the fundamental mode $E_{11}^x$of the input single-mode waveguide through the grating coupler. This mode is then coupled into a two-mode waveguide which supports two lateral modes ($E_{11}^x$and$E_{12}^x$) through a discontinuous junction formed by offsetting two waveguides with a distance of s. Due to the lateral asymmetry of the junction, the light excites these two lateral modes and the radiation modes simultaneously. Both $E_{11}^x$mode and $E_{12}^x$ mode propagate in the spiral sensing waveguide with different velocities, accumulating different phase changes and interfering with each other. Through another discontinuous junction at the exit of the sensing area, light in these two modes recombines and goes back into the fundamental mode of the output single-mode waveguide. Finally, the light is coupled out of the device to a photodetector for measurement through the output grating coupler. Since the mode indices of the two modes are affected differently by the RI change on the waveguide surface, their accumulated phase changes are different and the output light intensity is modulated by their net phase difference change. Therefore by measuring the output light intensity change, the external RI change can be tracked.

 

Fig. 1 (a) Schematic diagram of the proposed sensor. (b) Cross section of the single-mode waveguide. (c) Cross section of the two-mode waveguide and electric field intensity profiles for the two lateral modes.

Download Full Size | PDF

We form this sensor in silicon waveguide since it allows small bending radius which is ideal for compact device. Figure 1(b) shows the waveguide cross section of the single-mode waveguide with a silicon core on a SiO₂ substrate and covered with a SiO₂ cladding. The silicon core is W₁ wide and h high. The refractive indices of Si and SiO₂ are n_si and n_sio2, respectively. Figure 1(c) shows the cross section of the two-mode waveguide in the sensing area which is similar to Fig. 1(b) except that the cladding layer is replaced with the analyte to be measured (with a refractive index of n_ex) and the core width is increased to W₂ to support two lateral modes. The electric field intensity profiles for the $E_{11}^x$ mode and $E_{12}^x$ mode supported by the waveguide are also shown in Fig. 1(c). The sensing area is circumscribed by the black rectangle in Fig. 1(a) and the smallest radius of curvature of its spiral layout is D₀/2 and the spacing between two adjacent waveguides is d (be large enough to prevent evanescent coupling). The spiral is composed of two curved waveguides of (N + 1/4) turns having a linear variation of their radius of curvature, $R(\theta )=D_0+(d+W_2)\theta /\text{π}$, that are connected by two half-circles making an “S” section with a constant radius of curvature of D₀/2.

The power excitation ratio at the discontinuous junction affects the interference visibility of the modulated output signal and is decided by the geometries of both waveguides and offset distance s. Large visibility, i.e., large contrast of interference pattern, is preferred for good measurement accuracy. When optical power in the two modes equals to each other, the contrast becomes maximum. Figure 2(a) shows the dependence of the normalized excited power in $E_{11}^x$ and $E_{12}^x$ modes on the offset distance s for discontinuous junction with W₁ = 450 nm and W₂ = 900 nm. The parameters used in the calculations are n_Si = 3.48, n_SiO2 = 1.45, and h = 220 nm. The wavelength is 1550 nm. The excited power of each mode in the two-mode waveguide is analyzed with the finite difference time domain method (FDTD software package from Lumerical Solutions, Inc.). The offset distance s is actually an effective means for the control of the excitation ratio. As noted from Fig. 2(a), the excited power in $E_{12}^x$ mode can be equal to or larger than that in the$E_{11}^x$mode when s is larger than 485 nm. The total power loss at the discontinuous junction due to radiation mode excitation is ~10% (~0.5 dB), when s = 485 nm. The use of lateral modes instead of transversal modes [15,16] has the advantage of releasing the requirement of multiple lithography and etching steps, i.e., single-mode waveguide, two-mode waveguide and discontinuous junction are fabricated in one step.

 

Fig. 2 (a) Dependence of the normalized excited power in $E_{11}^x$ mode, $E_{12}^x$ mode, and total excited power in two modes on the offset distance s for a discontinuous junction with W₁ = 450 nm and W₂ = 900 nm. (b) Dependence of bending loss for $E_{11}^x$ and $E_{12}^x$ mode on the bending radius for two-mode waveguides with different widths W₂.

Download Full Size | PDF

Since the sensitivity of an MZI sensor can be enhanced by increasing the sensing waveguide length. We use a folded two-mode waveguide in spiral layout for achieving high sensitivity of long sensing arm while still has compact sensing area compatible with spotting machine and requiring small volume of sample. For a two-mode waveguide in spiral layout, the bending losses of the two modes are different. The light confinement of the $E_{12}^x$ mode is worse than $E_{11}^x$ mode and therefore its bending loss is higher. Figure 2(b) shows the bending losses for these two modes in two-mode waveguides with different widths: W₂ = 700 nm, 800 nm, 900 nm and 1000 nm. Waveguides with these four widths only support two lateral modes. The bending loss is analyzed with a full-vectorial finite difference mode solver (Mode solver software package from Lumerical Solutions, Inc.) and the refractive index of the analyte is assumed to be n_ex = 1.3105. As shown in Fig. 2(b), the bending loss of $E_{12}^x$ mode is several orders of magnitude higher than that of $E_{11}^x$ mode. Both modes in a two-mode waveguide with larger width have smaller bending loss due to better light confinement. As verified by the following studies, use of two-mode waveguide with narrower width can achieve higher sensitivity. However, the bending losses of these two modes and their difference also become larger as shown in Fig. 2(b) and result in poor visibility. This can be solved by using large bending radius where the bending loss is small enough. As shown in Fig. 2(b), the bending losses for both modes become negligible when the bending radius is larger than 30 μm. In practice, to achieve good visibility, the scattering loss induced by the non-smooth fabricated waveguide sidewall should be considered and handled carefully, since large scattering losses (especially the $E_{12}^x$ mode) lower the visibility and signal-to-noise ratio (SNR) of the output signal.

3. Device fabrication and characterization

3.1 Fabrication

The fabrication of the sensor chip started with a commercially available 200 mm silicon-on-insulator (SOI) wafer with 220 nm-thick top silicon layer and 2 μm-thick buried oxide (BOX) layers. First, the sensor structure was patterned by 248 nm deep UV lithography and etched to BOX by reactive ion etching (RIE) process. Next, input and output grating couplers with etch-depth of 70 nm and grating period of 630 nm were patterned and etched. Then a 2.9-μm SiO₂ overcladding layer was deposited on the etched structure using plasma enhanced chemical vapor deposition (PECVD). Finally, the sensing area was opened for the sensing area by dry and wet etching respectively. Figure 3(a) and (b) show scanning electron microscope (SEM) images of the fabricated spiral two-mode spiral waveguide and its central area (for spiral layout with D₀ = 30 μm, N = 5, d = 5 μm, W₂ = 900 nm). The microscopic image of the sensing area with the overcladding etched away is shown in Fig. 3(c). The total length of the two-mode waveguide in the sensing area was 4582 μm long and folded within a 185 μm diameter spot. To study the effect of the offset distance on the interference visibility, sensors with five different values of offset distance s were fabricated. Figure 4 shows the SEM images for discontinuous junctions with three different offset distance (s = 400 nm, 485 nm and 550 nm) before SiO₂ deposition. As observed in Fig. 4, the sharp corners at the junctions were smeared out to roundish shape but different offsetting effects are clearly seen. Sensors with identical spiral layout but four different sets of single-mode and two-mode waveguide width were fabricated: W₁/W₂ = 350/700 nm, 400/800 nm, 450/900 nm and 500/1000 nm.

 

Fig. 3 SEM images of fabricated (a) spiral two-mode waveguide and (b) its central area. (c) Microscopic image of the sensing area with the SiO₂ overcladding layer etched away.

Download Full Size | PDF

 

Fig. 4 SEM images of discontinuous junctions with offset distance (a) s = 400 nm, (b) s = 485 nm, and (c) s = 550 nm.

Download Full Size | PDF

3.2 Interference visibility

To characterize the effect of the offset distance on the interference visibility, transmission spectra of sensors with W₁/W₂ = 450/900 nm and five different values of offset distance s were measured with an amplified spontaneous emission (ASE) broadband light source and an optical spectrum analyzer (OSA). The two-mode sensing waveguides in these five sensors were 4500-μm long straight waveguides and therefore bending loss does not contribute to the total loss. Deionized (DI) water was applied on the sensing area. The transmission spectra are shown in Fig. 5, from where we clearly see the interference patterns. It is noted from Fig. 5 that the contrast of interference pattern increases as the offset distance s increases and reaches maximum (>15 dB) when s = 550 nm. According to the simulation results shown in Fig. 2(a), the maximum contrast is achieved at s = 485 nm for sensors with W₁/W₂ = 450/900 nm. The discrepancy is mainly attributed to the scattering loss that has not been considered in the simulation. Since the scattering loss of the $E_{12}^x$ mode is generally larger than the $E_{11}^x$ mode (stronger electric field at the core-cladding boundary), larger offset is needed to excite more optical power in the $E_{12}^x$ mode. Based on the results shown in Fig. 2(a) and Fig. 5, the propagation loss of the $E_{12}^x$ mode is estimated to be ~14% larger than that of the $E_{11}^x$ mode. The spectrum of the corresponding sensor with two-mode spiral sensing waveguide of similar length (D₀ = 30 μm, N = 5, d = 5 μm, and total length = 4582 μm) was also measured and an excess transmission loss of ~1.2 dB was found. Sensors with the other three sets of W₁/W₂ (350/700 nm 400/800 nm and 500/1000 nm) were also measured and interference contrast as large as 10 dB was obtained for both W₁/W₂ = 400/800 nm and 500/1000 nm at s = 550 and 600 nm, respectively. However, for sensors with W₁/W₂ = 350/700 nm, no clean interference pattern was observed for different offset distance from 350 to 550 nm, which was probably due to the large propagation loss of the $E_{12}^x$ mode and light in it was almost completely lost before reaching the output. The bulk and surface sensitivity of the sensor with W₁/W₂ = 450/900 nm, s = 550 nm, D₀ = 30 μm, and N = 5 was characterized in detail and the results are shown in the following sections.

 

Fig. 5 Transmission spectra for sensors with W₁ = 450 nm and W₂ = 900 nm at five different values of offset distance s.

Download Full Size | PDF

3.2 Bulk sensitivity

The bulk sensitivity of this sensor at wavelength λ is defined as the phase difference change between the two modes (for simplicity, we refer to it as “phase change”) per refractive index unit (RIU) change, i.e.,

For the phase change measurement, TE polarized light from a tunable laser and a polarization controller at wavelength 1550 nm with an output power of 3.2 mW was coupled into the sensor through the input grating coupler. The light at the output grating coupler was simultaneously monitored with an optical power meter which was remotely controlled through a computer interface developed in our laboratory using the LabVIEW program. The optical power reading versus time was recorded every 50 milliseconds. The phase change was directly extracted from the output intensity data by fitting it with a cosine function with a varying phase.

The bulk refractive index sensitivity of the sensor was characterized by injecting NaCl solutions of different concentrations (1, 3, 5, 7 and 10 w/w%) over the sensor. For flow control, an acrylic well designed for the attachment of Tygon tubes was adhered onto the surface of the chip prior to applying NaCl solution. The volume of the solution chamber was 32 μl. Liquid flow was controlled by a peristaltic pump. All experiments were carried out at room temperature. DI water was first injected into the liquid chamber at a flow rate of 8.5 µL/min and a reference optical power baseline was taken. After the baseline was stabilized, the NaCl solution of different concentrations was flown onto the sensor respectively. The mode indices of the two modes increased differently as NaCl solution was introduced, resulting in a net phase difference change between them. Therefore, the output optical intensity at a fixed wavelength changed accordingly as shown in Fig. 6(a). Figure 6(b) shows a linear increase of the phase change with increasing refractive index of the NaCl solution. A linear fit to the data sets shows bulk refractive index sensitivity of 461.6 π/RIU (R² = 0.9990) with ~4582 μm long two-lateral-mode sensing waveguide folded within a 185 μm diameter spot. The refractive index of the NaCl solution with p% concentration is obtained by $\text{n}(\text{p}\%)=1.3105+0.17151\times \text{p}\%$ [17]. The detection limit of a sensor is defined by

 

Fig. 6 (a) Real-time optical response for applying NaCl solution with different concentrations. (b) Variation of phase change with refractive index of NaCl solution. Linear fitting shows bulk sensitivity of ~461.6 π/RIU (R² = 0.9990).

Download Full Size | PDF

Figure 7(a) shows the measured and calculated bulk sensitivity of sensors with different two-mode waveguide widths (W₁/W₂ = 400/800 nm, 450/900 nm and 500/1000 nm). The experimental and simulation results agree well with each other. For sensors with identical spiral sensing waveguide length, sensing waveguide with narrower width gives higher bulk sensitivity. This is attributed to its larger sensitivity difference between the two modes (S_d in Eq. (1)), as verified by the calculated results shown in Fig. 7(b).

 

Fig. 7 (a) Bulk sensitivity of sensors with different two-mode waveguide widths (solid line: theory; solid square: experiment). (b) Dependence of S_d on the two-mode waveguide width W₂.

Download Full Size | PDF

3.3 Surface sensitivity

The surface senstivity of the sensor was investigated using the deposition of a polyelectroltye multilayer (PEM). PEMs are fabricated by alternating deposition of positively charged and negatively charged polymers and they are known to have well-controlled and reproducible thicknesses [18]. Before PEM deposition, the chip went through the following series of surface treatment. The chip was first treated with oxygen plasma. It was then immersed in a solution of 2% APTES (3-aminopropyltriethoxysilane) in a mixture of ethanol/H₂O (95%/5%, v/v) for 2 hours, followed by thorough rinsing with ethanol and DI water. It was next dried under a nitrogen stream and heated at 120 °C for 15 minutes.

In the experiment, a PEM film was built by alternately flowing the chamber with aqueous solutions of PSS (poly (sodium-4-styrenesulfonate), 1.0 mg/mL in 50 mM NaCl), DI water and PAH (poly (allylamine hydrochloride), 1.0 mg/mL in 50 mM NaCl) for 15 minutes each. First, DI water was applied and after the baseline is stabilized the solution of negatively charged PSS was flown onto the positively charged APTES modified sensor surface. Then DI water was applied to wash away the nonattached PSS. Positively charged PAH was next flown onto the PSS, followed by water washing again. One bilayer was thereby formed. The real-time optical response for the deposition of one bilayer of PSS/PAH is shown in Fig. 8(a). When DI water was applied, two discontinuities were observed in the optical response which indicated reversed phase change caused by washing of the nonattached PSS or PAH. The variation of the phase change with the number of bilayers (n) after the deposition of (PSS/PAH)_n is shown in Fig. 8(b). The linear fitting shows sensitivity of ~2.27 π/bilayer (R² = 0.9956). The sensitivity for surface mass S_m detection is given by:

 

Fig. 8 (a) Real-time optical response for the deposition of one bilayer of PSS/PAH. (b) Variation of the phase change with the number of bilayers (n) after the deposition of (PSS/PAH)_n. Linear fitting shows sensitivity of ~2.27 π/bilayer (R² = 0.9956).

Download Full Size | PDF

4. Biotin–streptavidin binding

To demonstrate the biosensing capability of the sensor, a well-characterized biotin–streptavidin interaction was used as a model system. The interaction between biotin and streptavidin is one of the highest non-covalent affinities (K_d = 10⁻¹³M) and is therefore stable and specific [20]. The APTES-treated sensor surface was first incubated with 1 mg/ml NHS-biotin in DI water for 1 hour. Pure PBS (phosphate buffered saline) was then flown into the chamber. After an optical power baseline was taken, solutions of streptavidin in PBS of varying concentrations were applied. Before the next concentration was introduced, the chamber was washed with PBS to remove the unbound streptavidin. Measurement for each concentration of streptavidin solution was typically 25 minutes. The real-time optical response to the adsorption of 10 μg/ml streptavidin to the biotinylated sensor surface is shown in Fig. 9(a), which depicts a total phase change of 5 π with an interaction length of 4582 μm. Figure 9(b) shows the phase change induced by streptavidin binding as a function of the concentration of streptavidin solution for a range of concentrations from 10 ng/ml to 10 μg/ml. As shown in Fig. 9(b), phase change increases as the concentration of streptavidin increases and is well linearly fitted on a log–log scale within the concentration range of the experiments (R² = 0.9708). The lowest concentration used in the experiment was 10 ng/ml, causing 0.033π phase change.

 

Fig. 9 (a) Real-time optical response to the adsorption of 10 μg/ml streptavidin to the biotinylated sensor surface. (b) Variation of phase change with different concentrations of streptavidin solution (plotted in log–log scale, R² = 0.9708).

Download Full Size | PDF

5. Discussion and conclusion

The demonstrated bulk sensitivity and minimum measurable concentration of streptavidin in this paper is not as good as our recently reported highly sensitive two-channel MZI sensor based on slot-waveguides [9]. However, according to Eq. (1), the sensitivity of our proposed sensor can be further enhanced by increasing the two-mode sensing waveguide length while keeping the sensing area compact. For example, as the length for a sensing waveguide with W₂ = 800 nm is increased to 12050 μm, the bulk sensitivity becomes 1972 π/RIU which is higher than that demonstrated in [9] (1864 π/RIU). Meanwhile, such a long two-mode waveguide can be folded within a compact spot area with a diameter of 300 μm (D₀ = 30 μm and d = 5 μm). Although longer spiral two-mode sensing waveguide gives higher sensitivity, long waveguide results in large propagation loss for both modes (especially the higher order mode) and therefore poor interference visibility of the sensor. The propagation loss mainly comes from the scattering loss which can be significantly reduced by reducing the sidewall roughness with smoothing technologies. For example, the loss of a 500 nm wide single-mode silicon waveguide can be reduced from 32 dB/cm by the conventional fabrication method to 0.8 dB/cm with the oxidation smoothing technique [21]. Therefore, it is possible to increase the sensitivity further using long sensing waveguide with small scattering loss. In addition, as seen from Eq. (1), increase of the sensitivity difference between the two modes also helps to increase the sensor sensitivity, which can be achieved by the use of narrow two-mode waveguide as verified by the experimental results shown in Fig. 7(a). However, for waveguide as narrow as 700 nm, there is more stringent requirement on the fabrication for low propagation loss since the higher order mode in this waveguide is near cutoff. Finally, although sensors of spiral layout are demonstrated in this paper, the single sensing channel of this sensor can be folded into layouts of various shapes for certain specific applications (as long as the minimum bending radius is appropriately selected), where the layout of the sensing element has to been aligned with the microfluidic channel.

In summary, we have proposed and demonstrated a compact single-channel Mach-Zehnder interferometric biochemical sensor based on the use of two-lateral-mode spiral waveguide for sensing and discontinuous junction for mode excitation. The effect of discontinuity offset distance on the interference visibility has been studied. The bulk and surface sensitivity of fabricated sensors are characterized and the biosensing capability of the sensor is verified by the measurement of biotin–streptavidin interaction of different concentrations. The sensor can offer both high sensitivity and a compact sensing area which is ideal for the detection of small sample volume with an ultra-low detection limit.