By Topic

• Abstract

SECTION 1

## INTRODUCTION

Applications in the life sciences, drug discovery, medical diagnostics, environmental monitoring, food safety, as well as protection against biological and chemical warfare [1], [3], [3] are driving increasing demand for biochemical sensors. Such sensors should be highly sensitive, label free, compact, and easy to use. As attractive alternatives to several technologies employing nanowire [4] and nanopore [5] conductance measurements, nanowire crossbar arrays [6], nanoparticle probes [7], and nanomechanical resonators [8], optical waveguides are a promising technology to fulfill these requirements. Optical-waveguide-based approaches work by sensing the presence of analytes as effective refractive index (RI) changes displayed by the guided mode, and thus avoiding prior fluorescent or radioactive labeling [9]. Furthermore, optical waveguides can be integrated with electronic and fluidic components to form lab-on-a-chip bioanalytical microsystems [10]. Some examples of integrated optical biosensors include sensing elements based on surface plasmon resonance (SPR) [11], directional couplers [12], Mach–Zehnder [13] or Young's interferometers [14], Bragg grating [15] and optical microcavities, such as microtoroids [16], microrings [17] or photonic crystals [18]. High-quality factor optical microcavities can provide a large equivalent light-analyte interaction length with a small device size, which also significantly reduces the minimum amount of analytes required for detection, i.e., the detection limit [19].

The sensitivity of an optical waveguide sensor relies on the amount of light in the medium to be sensed. Recently, the slot waveguide was proposed as a guiding structure capable of high sensitivity in biochemical sensing applications [20]. The waveguide has two strips of high-RI material separated by a slot and supports a single quasi-transverse electric (TE) mode highly confined in the low-index submicrometer gap between the two strips [21], [22]. New chemically amplified electron-beam resists applied using electron-beam lithography makes it possible to create a single-mode waveguide with multiple slots to enhance the interaction between the light and the analyte. The work presented here theoretically and experimentally demonstrates that the increased number of slots and the strong electric field in the slots significantly increase the sensitivity of both homogeneous sensing and surface sensing.

SECTION 2

## SIMULATIONS OF MULTI-SLOT WAVEGUIDES

A schematic diagram of the multi-slotted waveguide is shown in Fig. 1. The modal profile and effective index of this guiding structure was analyzed with a fully vectorial 2-D mode solver FIMMWAVE (Photon Design), which is based on a film mode matching (FMM) algorithm [23]. This method is effective with modes that are near cutoff for structures composed of very thin layers. These features make the solver especially suited for modeling the proposed multi-slot waveguide for biosensing applications. The RI of SU8 polymer (Microchem) is 1.565 at the optical communications wavelengths, and that of silicon dioxide is 1.45. We also assumed the solution covering the waveguide is water (RI 1.33). As shown in Fig. 2, the modal profiles have peaks in the slots for the TE polarizations, and have dips for the transverse magnetic (TM) polarizations.

Fig. 1. Structure of the multiple-slot waveguide. (a) 3-D diagram of the waveguide structure. (b) Scanning electron micrograph of the fabricated SU8 waveguide.
Fig. 2. Modal profile of the multiple-slot waveguide. (a) and (b) Simulated field distribution (upper) and energy density (lower) for the TE (left) and TM (right) polarizations, respectively.

If used for detection of a homogeneous solution, the RI of the solution changes with analyte concentration and thus varies the effective index of the guided modes. The waveguide sensitivities for homogeneous sensing were numerically estimated by varying the analyte solution RI from 1.33 to 1.331 and finding the relevant change of effective index, i.e., $S_{wh} = \Delta n_{eff}/\Delta n_{s}$. From the variational theorem for dielectric waveguides, it is known that this sensitivity is proportionally related to the optical field confinement factor in the solution region [24]. Because multiple slots allow more light to be confined in the solution, it can intuitively be expected that the sensitivity would increase with number of slots (until the waveguide becomes leaky). It can also be expected that the TE mode has higher sensitivity than the TM mode. This is verified by our simulations with FIMMWAVE [see Fig. 3(a)].

Fig. 3. Simulation results of waveguide sensitivities as a function of the number of slots in the waveguide structure shown in Fig. 1(a). The polymer waveguides were divided into strips with equal widths. To highlight the effect of the increased number of slots, other waveguide parameters were kept constant. The slot width is fixed at 200 nm. For this waveguide structure, no guiding mode was found for a slot number greater than that for (a) homogeneous sensing and (b) surface sensing.

As for surface sensing, the optical waveguide guided modes are perturbed by the adsorption or binding events of bio-molecules at the surface of the waveguides. We modeled these adsorbed analyte molecules with a thin ad-layer over all the exposed area of the waveguides. The ad-layer thickness and RI were assumed to be 5 nm and 1.5, which are reasonable approximations of an adsorbed protein layer [25]. Changing the ad-layer thickness from 0 to 5 nm, the waveguide sensitivities for surface sensing were estimated by the quotient of effective index and ad-layer thickness differences, i.e., $S_{ws} = \Delta n_{eff}/\Delta t$. Our simulation shows that the waveguide sensitivity indeed increases with the number of slots [see Fig. 3(b)]. This result is also expected by the variational theorem, which tells us that the waveguide sensitivity is proportional to the optical power confined in the ad-layer cross section [24], considering larger ad-layer cross section area for more slots in the waveguide. Of course, the relation between waveguide sensitivity and ad-layer cross section area is not linear due to non-uniform optical field distribution in the ad-layer.

Even higher sensitivities for homogeneous or surface sensing are possible by optimizing other waveguide parameters, including slot width, strip width, waveguide height, etc. as more slots are incorporated in a single mode waveguide. In theory there is no upper limit to the number of slots. In practice it is limited by the device fabrication techniques and the ability for analytes to diffuse into the slots for detection. We can always expect higher sensitivities with more slots. As more slots are introduced, the effective index decreases, and single mode waveguides can be realized with larger waveguide width and height. This is advantageous in two aspects. A large waveguide mode makes it easier to couple light in and out of the waveguide. Larger waveguide width also further increases the surface area. Our simulation with a waveguide of 8 $\mu\hbox{m}$ in height and 4 $\mu\hbox{m}$ in width divided by 39 slots of 100 nm width shows that this waveguide only supports a TM mode (see Fig. 4). The surface area of this waveguide is around 107 times higher than that of a $2\ \mu\hbox{m} \times 2\ \mu\hbox{m}$ waveguide and sensitivity enhancement for surface sensing of similar scale is possible. The waveguide sensitivity for homogeneous sensing $S_{wh}$ is 0.411, around 1.5 times higher than that of the waveguide shown in Fig. 1. Simulations also show that it is possible to get even higher sensitivities with a multiple slot waveguide made of some high-RI-contrast materials, such as silicon on insulator (SOI). The multiple-slot SOI waveguide in Fig. 5 has a $S_{wh}$ of 0.67 for the TE polarization.

Fig. 4. Simulated modal profile (intensity distribution) for TM polarization of a SU8 multiple-slot waveguide. The cover medium is water and the substrate is $\hbox{SiO}_{2}$. The waveguide height is 8 $\mu\hbox{m}$. The waveguide is divided into 40 strips with widths of 100 nm by 39 slots with width of 100 nm.
Fig. 5. Simulated modal profile (intensity distribution) of a silicon (RI 3.48) multiple-slot waveguide. The cover medium is water and the substrate is $\hbox{SiO}_{2}$. The waveguide height is 500 nm. The waveguide is divided into 10 strips with widths of 30 nm by 9 slots with width of 30 nm. (a) TE polarization has higher intensity in the slots. (b) TM polarization has lower intensity in the slots.
SECTION 3

## DEVICE DESIGN AND FABRICATION

To convert the effective index change induced by the analyte solution concentration to measurable signals, the multi-slot waveguide structure was applied to the basic microring resonator device geometry with one bus waveguide. Microring resonators of the conventional strip waveguide geometry (no slots) were also fabricated to serve as references. The multi-slot waveguide structure is the same as Fig. 1(a), i.e., four strips that are 2 $\mu\hbox{m}$ high and 500 nm wide, separated by three slots that are 200 nm wide. The strip waveguide cross section is $2\ \mu\hbox{m} \times 2\ \mu\hbox{m}$. The bus waveguide to ring resonator coupling gaps are 200 nm for both groups of microring resonators because our simulations show that for the multi-slot waveguides, the highest coupling efficiency is achieved when the coupling gap width is the same as the slot width. The ring resonators have a racetrack shape. The radius of the curved sections is 200 $\mu \hbox{m}$. Devices with coupling lengths of $0 \sim 100\ \mu \hbox{m}$ for both waveguide structures have been fabricated to find out the optimal design [see Fig. 6(a)].

Fig. 6. Measurements of the multi-slot waveguide microring resonators fabricated with SU8 on the silicon-oxide-coated silicon substrate. (a) Excitation of the device with 636-nm laser through fiber coupling. (b) Measured transmission spectra of the device covered with DI water as the top cladding.

All devices were fabricated with SU8 polymer on a silicon substrate. Polymer materials are low cost and ease of fabrication is high. The low RI contrast of polymer waveguides also provides low surface scattering loss and high coupling efficiency to optical fibers. As a commonly used negative resist, SU8 can be directly patterned with an electron beam or ultraviolet (UV) light. Furthermore, SU8 is biocompatible [26] and bio-molecules can be directly immobilized to the surface of SU8 through hydrophobic adsorption and without prior silanization [27], which is advantageous for functionalizing the devices to provide specific sensing.

We used a standard electron beam lithography process to define the SU-8 waveguide structures. First, a 2- $\mu\hbox{m}$ thick SU-8 film was spin coated on a Si substrate, with $\sim\! 5\hbox{-}\mu\hbox{m}$ thermal oxide serving as the lower cladding. Then, an FEI Sirion scanning electron microscope of 30-kV accelerating voltage equipped with a Nanometer Pattern Generation System (NPGS) was used to generate the designs of waveguides and devices. The exposed regions were developed with an SU-8 developer. The fabrication resolution was smaller than 100 nm. The waveguide sidewall roughness was directly defined by the lithography process, although some post-fabrication technique, like thermal reflow, could have been used to reduce the roughness.

SECTION 4

## BIOCHEMICAL SENSING EXPERIMENTS

For the transmission spectrum measurement, individual microring resonators were cleaved from the Si wafers. The output of an erbium amplified spontaneous emission (ASE) broadband source with a wavelength range from 1520 to 1560 nm was polarized through an Agilent 8169A polarization controller, which consists of individually rotatable linear polarizer, half-wave plate, and quarter-wave plate and can synthesize any predetermined state of polarization. TM or TE light was respectively end coupled to one port of the bus waveguide through a polarization maintaining (PM) optical fiber. The output port was fiber coupled to an optical spectrum analyzer (HP 70951B) to obtain the resonance spectrum.

The devices with zero coupling lengths (circular rings) presented the highest resonance extinctions for both waveguide structures and are close to the critical coupling condition [see Fig. 6(b)]. In the following experiments, only circular ring resonator devices were used to compare the sensing capabilities of the two guiding structures. The multi-slot waveguide ring resonators offered quality factors of around 7000, and the conventional strip waveguide achieved quality factors $(Q)$ of more than 15 000. The corresponding waveguide losses are 16 dB/cm and 4 dB/cm. The higher loss of the multi-slot waveguide should be attributed to scattering due to imperfections in the waveguide surfaces. Introducing more slots in the waveguide structure makes the waveguide loss more sensitive to side wall smoothness. Although the higher loss and reduced $Q$ do not affect the amount of resonance shift and sensitivity, they may broaden the resonances and reduce measurement accuracy. However, in our experiments, resonance broadening was not found to be significant enough to seriously impact the accuracy of resonant wavelength measurements.

For microring resonator devices, the effective index variation shifts the resonant wavelengths. Thus the analyte concentration in the solution can be detected by monitoring the shift of the resonant wavelength. Sodium chloride (NaCl) solutions in DI water with mass concentrations 0 $\sim$ 4% were used to determine the sensitivity of the microring resonators. The RI of a NaCl aqueous solution changes 0.0018 RIU (RI unit) per 1% mass concentration at 20 °C [28]. Plots of the resonant wavelength versus NaCl solution concentration (and the solution RI) present linear relationships for both ring resonators [see Fig. 7(a)]. The detection sensitivity is obtained from the slopes of the plots, and the sensitivity of the multi-slot waveguide ring resonator was calculated through line fitting to be 244 nm/RIU, which is nearly five times the sensitivity of the conventional strip waveguide ring resonator (55 nm/RIU). This enhancement stems from the fact that for similar resonant wavelengths, $\lambda_{r}$, and $n_{eff}$, the device sensitivity [29], $S_{hr} = \partial \lambda_{r}/\partial n_{s} = \partial \lambda_{r}/\partial n_{eff}\cdot S_{wh} = \lambda_{r}/n_{eff}\cdot S_{wh}$, is defined by the waveguide sensitivities $S_{wh}$. Assuming $\lambda_{r} = 1550\ \hbox{nm}$, $n_{eff} = 1.5$, the waveguide sensitivity $S_{ws}$ for the multi-slot waveguide and the conventional strip waveguide are 0.24 and 0.05, respectively, which agrees very well with our simulations in Fig. 3(a).

Fig. 7. Comparative biochemical sensing experiments for the microring resonators of two different waveguide structures. (a) Resonant wavelength shift as functions of the NaCl solution concentration and refractive index. (b) Resonant wavelength shift over different surface processing steps.

Alternatively to monitoring of the resonant wavelength shift, the guided mode effective index change can be also deduced from output intensity variation at the wavelength with the maximum slope in the transmission spectrum. For this sensing scheme, the device sensitivity becomes $S_{hi} = \partial I_{o}/ \partial n_{s} = I_{i}\cdot \partial T/\partial n_{eff}\cdot S_{wh} = I_{i}\cdot Q/n_{eff}\cdot S_{wh}$, where $I_{o}$ and $I_{i}$ are the output intensity and the input intensity, and $T = I_{o}/I_{i}$ is the transmittance. The advantage of this approach is that detection limit may be orders lower than measuring the resonance spectrum. The disadvantage is that intensity measurement is sensitive to the coupling condition and noise.

A microring resonator with multi-slot waveguide and a microring resonator with unslotted strip waveguide were compared for surface sensing. Both resonators were fabricated from SU8 polymer of the same thickness. The two microring resonators were first rinsed with deionized (DI) water 3 times and the resonant spectra were measured as references. The devices were then immersed in a solution of 2 mg/ml biotinylated bovine serum albumin (biotin-BSA, Pierce Biotechnology) in phosphate buffered saline (PBS, pH 7.2, Sigma-Aldrich) solution and incubated for 30 min. Because of the hydrophobic nature of cured SU8, the protein molecules were immobilized to the waveguide surface through physical adsorption and formed a monolayer [27], [30]. The biotin-BSA ad-layer, free biotin-BSA molecules and saline substances in the buffer solution shifted the resonant peaks toward longer wavelengths for both ring resonators. Subsequently, the un-adsorbed biotin-BSA and saline substances were removed by rinsing the devices with DI water again for 3 times, leading to reverse shifts of the resonant wavelengths for both ring resonators. The net red shift is 0.51 nm for the multi-slot waveguide ring resonator and 0.15 nm for the conventional strip waveguide ring resonator [see Fig. 7(b)]. These net shifts are due to the adsorbed biotin-BSA monolayer on the waveguide surfaces.

The device sensitivity for surface sensing can be defined as the resonant wavelength shift with the ad-layer thickness change [29], i.e., $S_{sr} = \partial \lambda_{r}/\partial t = \partial \lambda_{r}/\partial n_{eff}\cdot S_{ws} = \lambda_{r}/n_{eff}\cdot S_{ws}$. The size and molecular weight of biotin-BSA molecule are about 4 nm × 4 nm × 14 nm and 66 432 Da [17]. Supposing $\lambda_{r} = 1550\ \hbox{nm}$, $n_{eff} = 1.5$, the corresponding waveguide sensitivity $S_{ws}$ for the multi-slot waveguide and the conventional strip waveguide are $1.2 \times 10^{-4}\ \hbox{RIU/nm}$ and $3.6 \times 10^{-5}\ \hbox{RIU/nm}$, which are consistent with our simulated results shown in Fig. 3(b). The waveguide sensitivity of the conventional strip waveguide is close to (a little higher than) a simulated one of the polymer slab waveguide [31]. The sensitivity of the multi-slot device is on the same order of current commercial SPR biosensors [11], [32]. However, microring resonator sensors are much compact and use less analyte. These biotin-BSA ad-layers provide binding sites for ligands like avidin or streptavidin molecules, and BSA can block nonspecific adsorption of other proteins [30]. The bound avidin or streptavidin layer can be further used to detect other biotinylated bio-molecules.

SECTION 5

## CONCLUSION

We have studied a novel optical waveguide structure with multiple slots and demonstrated its homogeneous and surface sensing capabilities with the polymer microring resonator device architecture. This waveguide geometry results in much stronger optical field confinement in the submicrometer and lower RI slot regions, which allows highly sensitive and label-free biochemical sensing. By incorporating three slots in the waveguide, an increase in sensitivity of five times for homogeneous sensing and three times for surface sensing is demonstrated. Theoretically there is no limit to the number of slots in the waveguide. As nanolithography technology advances, a larger number of smaller slots can be fabricated in a single mode waveguide and the sensitivity gain will increase accordingly. Although our study was focused on vertically oriented slots, the same principle can be extended to waveguides with multiple horizontal air gaps. Such waveguides may be fabricated in multilayer thin films with alternating layers of two materials [33] and selective removal of the layers of one material by chemical etching. Simulations suggest that a 100-fold increase in the sensitivity of surface sensing can be expected. Our multiple-slot approach increases the sensitivity while maintaining strong confinement and reliable operation of the waveguide. The waveguide structure can be easily applied to other device architectures like directional couplers, Mach–Zehnder or Young's interferometers. More compact structures and higher sensitivity can be expected if microring resonators with multiple slots can be realized in the SOI material platform [31]. The strong evanescent wave interaction with the surrounding liquid medium (such as analyte solutions or laser dyes) makes it advantageous to integrate this type of waveguide with microfluidic channels to form complete lab-on-a-chip systems.

### ACKNOWLEDGMENT

The work was conducted at the Nanotech User Facility at the University of Washington: a member of the National Nanotechnology Infrastructure Network (NNIN) supported by the National Science Foundation.

## Footnotes

This work was supported by the National Science Foundation (NSF) under Grant ECS-0437920 and Grant NSF-DMR-0092380 and the NSF Center on Materials and Devices for Information Technology Research (CMDITR) under Grant DMR-0120967. Corresponding author: H. Sun (e-mail: hssun@u.washington.edu).

## References

No Data Available

## Cited By

No Data Available

None

## Multimedia

No Data Available
This paper appears in:
No Data Available
Issue Date:
No Data Available
On page(s):
No Data Available
ISSN:
None
INSPEC Accession Number:
None
Digital Object Identifier:
None
Date of Current Version:
No Data Available
Date of Original Publication:
No Data Available