Applications of molecular sensors in genomics, proteomics, and drug screening often require high throughput measurements of many different binding reactions. Therefore, the ability to form sensor arrays where molecular receptors can be independently immobilized on each array element is essential for new sensor technologies. The evanescent-field waveguide sensor has proven attractive for this application owing to its potential for high density integration, label-free operation, and real-time monitoring capability. Various waveguide sensors have been reported using a variety of material systems and configurations such as Mach–Zehnder interferometers (MZIs) [1, 2, 3, 4], microdisks [5, 6], and ring resonators [7, 8, 9, 10]. They can provide an increased sensitivity compared with established techniques such as surface plasmon resonance (SPR) [11] and may be multiplexed using established methods in integrated optics.

In previous work, we have developed highly sensitive discrete silicon-on-insulator (SOI) photonic wire waveguide sensors, and we have shown that their small size and high refractive index contrast produce an unusually strong evanescent field at the waveguide surface [3]. By implementing these waveguides in a balanced MZI configuration, high sensitivity to molecular binding was achieved while almost eliminating temperature induced sensor drift [4]. In this Letter, we report an integrated silicon biochip containing a multiplexed array of these sensors. The chip also incorporates a monolithically integrated microfluidic channel for sample delivery and our experiments are carried out using a commercial microarray spotter to deposit receptor molecules on individual sensor elements.

The response of an evanescent-field sensor to molecular binding at the sensor surface is proportional to $\partial N_{\text{eff}}/\partial D_{\text{opt}}$, where $\partial N_{\text{eff}}$ is the effective index change of the waveguide mode and $\partial D_{\text{opt}}$ is the change in optical thickness of the formed molecular layer. Since $\partial N_{\text{eff}}/\partial D_{\text{opt}}$ increases with the waveguide core-cladding refractive index contrast and with decreasing core thickness, thin SOI waveguides have a higher response to molecular binding than waveguides fabricated in most other materials [3]. In a SOI photonic wire waveguide, the transverse magnetic (TM-like) mode provides the highest sensitivity, which is maximized for a waveguide thickness of $\sim 0.22\mu \text{m}$ [3].

To transform the induced effective index change into a measurable intensity change, we have developed MZI sensors using $0.26\mu \text{m}\times 0.45\mu \text{m}$ photonic wire waveguides near this optimum silicon thickness. The MZI arms were designed to have similar optical path lengths to minimize the device response to temperature changes. The photonic wires were defined using electron beam lithography and reactive ion etching on a SOI wafer with a $0.26\text{-}\mu \text{m}$-thick silicon layer and a $2\mu \text{m}$ buried oxide.

The waveguides were covered with a $2\text{-}\mu \text{m}$-thick SU-8 polymer upper cladding for optical isolation of the waveguide mode from the analyte solution. The SU-8 was removed over the MZI sensing arms forming circular windows, thereby exposing the waveguides. Both the reference and sensing arms of the MZI contain a $\sim 1.8\text{- mm}$-long double spiral waveguide occupying a circular area of $130\mu \text{m}$ diameter. The cumulative phase shift in the MZI arm is proportional to both the surface sensitivity $\partial N_{\text{eff}}/\partial D_{\text{opt}}$ of the waveguide and its length [4]. The spiral waveguide layout offers the high sensitivity of a long waveguide, while providing a compact geometry for compatibility with microarray spotting. Furthermore, when placed in a typical microfluidic channel having a width comparable to the sensor diameter $\left(\sim 100–300\mu \text{m}\right)$, the spiral waveguide occupies a much larger fraction of the channel floor per unit channel length compared with a conventional straight waveguide for an improved molecular capture efficiency.

Sensors were multiplexed in a $400\mu \text{m}\times 400\mu \text{m}$ grid pattern as shown in Fig. 1 . At the input facet of the chip the nominal waveguide width of $0.45\mu \text{m}$ was reduced to $0.15\mu \text{m}$ using a linear taper for adiabatic expansion of the waveguide mode for efficient coupling with a lensed optical fiber. The measured fiber-to-waveguide coupling loss was found to be less than 3 dB, and the waveguide propagation loss was measured to be less than 0.6 dB/mm. Following the input waveguide, a $1\times 8$ optical power splitter consisting of cascaded $1\times 2$ y junctions was used to direct light to six sensors in the array (S1–S6 in Fig. 1) and two optical test ports. The lower three sensors shown in Fig. 1 are independent test structures and were not used for this sensor array experiment. The outputs of the six MZI sensors were directed to the output chip facet.

After defining the sensor windows, a second SU-8 layer of $50\mu \text{m}$ thickness was spun onto the chip. A $190\text{-}\mu \text{m}$-wide microfluidic channel was defined in this layer using contact lithography. It was aligned to pass over the SU-8 sensing windows as shown in Fig. 1. The channel is terminated at both ends with 1 mm diameter circular reservoirs. The reservoirs are added to increase the alignment tolerance of the chip to inlet and outlet fluidic connections contained in a polydimethylsiloxane (PDMS) cover plate, which seals the top of the channel during testing.

After defining the fluidic channel, an oxygen plasma was used to form a 3-nm-thick ${\text{SiO}}_2$ layer on the exposed waveguides. This allows standard glass-based functionalization procedures to be subsequently used for molecular attachment. After fabrication, the sensor surface was cleaned and activated using nitric acid and silanized with 3-aminopropyltriethoxysilane vapor in a vacuum chamber.

The sensor array was tested using binding reactions between complementary antibody–antigen pairs: goat and antigoat immunoglobulin G (IgG) and rabbit and antirabbit IgG. Each sensor element in the array was individually functionalized using a noncontact robotic spotting tool (GeSiM Nano-Plotter 2.1). The spotter was used to deposit drops of solutions containing the appropriate receptor molecule [diluted in phosphate-buffered saline (PBS)] on each exposed sensor element in the array. A 450 pl drop volume was chosen to fill the $130\mu \text{m}$ sensor diameter, and in each case the molecular concentration was 1 mg/ml. Goat IgG was deposited on sensors 1 and 2 (S1 and S2 in Fig. 1), rabbit IgG was deposited on sensors 3 and 4 (S3, S4), and bovine serum albumin (BSA) was deposited on sensors 5 and 6 (S5, S6). The spotted chip was then allowed to dry in 40% humidity at $25°\text{C}$. The PDMS cover plate was then aligned and adhered to the chip to seal the SU-8 channel. It was attached without the use of epoxies or other binding agents, and therefore could be removed and reused. A syringe pump was attached to the outlet port of the PDMS cover and was used to draw solutions through the fluidic channel. A microfluidic switch was attached to the inlet port to switch between different analyte and buffer solutions. After attaching the cover piece a wash step was performed where the PBS was passed through the channel for 10 min to remove any loosely bound receptor molecules and residue left from the spotting process.

A lensed optical fiber delivered a TM polarized 1560 nm laser signal to the input waveguide of the sensor array. The six output waveguides were simultaneously monitored with a high sensitivity $320\times 240\text{\hspace{0.17em} pixel}$ IR InGaAs camera. Intensity readings from a single pixel near the center of each imaged waveguide output were acquired every 200 ms using a 16 ms exposure time. A fiber coupled optical power of approximately 100 nW provided a sufficient intensity at each array output waveguide to nearly saturate the camera pixels at the MZI fringe maxima, thereby achieving a 36 dB dynamic range (i.e., 12 bit resolution). Therefore, the number of sensors that can be interrogated using this approach will not be practically limited by loss owing to optical power splitting when used with readily available laser sources with output powers of a few milliwatts.

Figure 2 shows the measured phase shift induced in the MZI sensing arm during the experiment. The phase shift was extracted from the raw MZI intensity data by fitting it with a cosine function with a varying phase. The phase shift was then cumulated at each time point and the original signal noise is preserved and transferred to the phase plots. After the rinse step (ending at $t=0$), the PBS continued to flow through the fluidic channel for 3 min, followed by antirabbit IgG (200 nM, from inoculated chicken) and then rinsed with the PBS. Sensors 3 and 4 show strong binding of the rabbit IgG receptors to the antirabbit IgG in solution, with total phase shifts of 2 $\left(2\pi \right)$ and 2.4 $\left(2\pi \right)$ after the rinsing step. As expected, sensors 1 and 2 functionalized with goat IgG demonstrate a much smaller (and $\sim 10$ times slower) response to the antirabbit IgG. We attribute their residual response to nonspecific binding of antirabbit IgG to the silanized sensor surface, which should be further reduced by implementing a blocking step prior to testing. Following exposure to the antirabbit IgG and rinse step, a solution of antigoat IgG (200 nM, from inoculated rabbit) was then passed through the channel. In this step, sensors 1 and 2 now show a strong binding reaction between the goat-IgG receptor and the antigoat IgG, with a rapid change in phase saturating at 2.6 and 2.3 $\left(2\pi \right)$. At the same time, sensors 3 and 4 demonstrate a response of less than 0.03 $\left(2\pi \right)$. The control sensors 5 and 6 were printed with the BSA, which is a commonly used blocking agent that should not bind to either molecule. The BSA effectively blocked these control sensors, which gave cumulative phase responses of less than 0.1 $\left(2\pi \right)$ during each anti-IgG step.

An average rms phase noise of $4\times {10}^{-4}$ $\left(2\pi \right)$ was observed on the MZI response curves. Based on a saturated phase shift for the antirabbit IgG of approximately 2 $\left(2\pi \right)$ this indicates that approximately 0.02% of a saturated molecular monolayer could be resolved. From ellipsometry measurements performed on a witness sample, which was simultaneously functionalized with the sensor chip and exposed to the same concentration of analyte, we estimate a surface mass coverage of $1.2\text{\hspace{0.17em} ng}/{\text{mm}}^2$ [12]. The measured phase noise therefore corresponds to a resolvable surface coverage of $\sim 0.25\text{\hspace{0.17em} pg}/{\text{mm}}^2$. Owing to the small surface area of the sensing waveguides, the resolvable surface coverage corresponds to a low molecular mass of $\sim 0.5\text{\hspace{0.17em} fg}$. These values compare favorably with those from commercial SPR systems, and we believe that our noise floor can be further improved by implementing multiple-pixel averaging in the InGaAs camera readout.

The measured temperature dependence of the MZI devices when exposed to the PBS solution was measured using a heated stage and was found to be less than 0.005 $\left(2\pi \right)/°\text{C}$. This value corresponds to $\sim 0.2\%$ of the signal induced by the formation of the anti-IgG layer per $°\text{C}$. This is sufficiently low for typical sensing applications to avoid the need for high-precision temperature control or reference sensors that may be required using other transducer configurations, such as resonators.

In summary, we have demonstrated a multiplexed biosensor array using silicon photonic wire waveguides to detect binding reactions at individually functionalized sensors. We have used simple optical power splitters to divide light from a single input waveguide to an MZI array and a high sensitivity InGaAs camera for parallel readout of sensors. Analyte was delivered to the device using a monolithically integrated microfluidic channel that greatly reduces alignment challenges with hybrid fluidic approaches.

 

Fig. 1 Top view of a silicon photonic wire MZI sensor array, with a monolithically integrated SU-8 microfluidic channel, that is aligned to the sensor windows for analyte delivery. The top six sensors (S1–S6) of the array are used in the experiment, while the bottom three sensors are independent test structures.

Download Full Size | PDF

 

Fig. 2 Binding curves obtained using an IR camera to simultaneously track the phase change induced in the MZI sensors. Sensors 1 and 2 were printed with goat IgG, sensors 3 and 4 were printed with rabbit IgG, and sensors 5 and 6 were printed with BSA using a microarray spotter. For clarity the sensor response curves have been offset in the plot.

Download Full Size | PDF

1. B. J. Luff, J. S. Wilkinson, J. Piehler, U. Hollenback, J. Ingenhoff, and N. Fabricius, J. Lightwave Technol. 16, 583 (1998). [CrossRef]  

2. B. Sepulveda, J. S. del Rio, M. Moreno, F. J. Blanco, K. Mayora, C. Dominguez, and L. M. Lechuga, J. Opt. A 8, S561 (2006). [CrossRef]  

3. A. Densmore, D. -X. Xu, P. Waldron, S. Janz, P. Cheben, J. Lapointe, A. Delâge, B. Lamontagne, J. H. Schmid, and E. Post, IEEE Photon. Technol. Lett. 18, 2520 (2006). [CrossRef]  

4. A. Densmore, D. -X. Xu, S. Janz, P. Waldron, T. Mischki, G. Lopinski, A. Delâge, J. Lapointe, P. Cheben, B. Lamontagne, and J. H. Schmid, Opt. Lett. 33, 596 (2008). [CrossRef]   [PubMed]  

5. A. M. Armani, R. P. Kulkarni, S. E. Fraser, R. C. Flagan, and K. J. Vahala, Science 317, 783 (2007). [CrossRef]   [PubMed]  

6. R. W. Boyd and J. E. Heebner, Appl. Opt. 40, 5742 (2001). [CrossRef]  

7. D. X. Xu, A. Densmore, A. Delâge, P. Waldron, R. McKinnon, S. Janz, J. Lapointe, G. Lopinski, T. Mischki, E. Post, P. Cheben, and J. H. Schmid, Opt. Express 16, 15137 (2008). [CrossRef]   [PubMed]  

8. K. De Vos, I. Bartolozzi, E. Schacht, P. Bienstman, and R. Baets, Opt. Express 15, 7610 (2007). [CrossRef]   [PubMed]  

9. A. Yalcin, K. C. Popat, J. C. Aldridge, T. A. Desai, J. Hryniewicz, N. Chbouki, B. E. Little, O. King, V. Van, S. Chu, D. Gill, M. Anthes-Washburn, M. S. Unlu, and B. B. Goldberg, IEEE J. Sel. Top. Quantum Electron. 12, 148 (2006). [CrossRef]  

10. A. Ksendzov and Y. Lin, Opt. Lett. 30, 3344 (2005). [CrossRef]  

11. W. Lukosz, Biosens. Bioelectron. 6, 215 (1991). [CrossRef]  

12. H. Elwing, Biomaterials 19, 397 (1998). [CrossRef]   [PubMed]