Simulation Evaluation of a Label-free Silicon-on-Insulator “Lab on a Chip” Optical Biosensor
Najala Najeeb1,2, Yaping Zhang1*, Trevor Benson2
1Department
of Electrical and Electronic Engineering, University of Nottingham, China
2Department of Electrical and Electronic Engineering, University of Nottingham, UK
*Corresponding author: Yaping Zhang, Department of Electrical and Electronic Engineering, University of Nottingham, 199 Taikang East Road, Ningbo, 315100, China. Tel: +8657488182420; Email: yaping.zhang@nottingham.edu.cn
Received
Date: 30 June, 2018; Accepted Date: 17 July, 2018; Published
Date: 23 July, 2018
Citation: Najeeb N, Zhang Y, Benson T (2018) Simulation Evaluation
of a Label-free Silicon-on-Insulator “Lab on a Chip” Optical Biosensor. Biosens
Bioelectron Open Acc: BBOA-140. DOI: 10.29011/ 2577-2260.100040
1. Abstract
This paper presents the design and optimization of a Silicon-on-Insulator (SOI) “Lab on a Chip” immunosensor, based on interferometer technology, capable of the label-free, real time, parallel detection and identification of multiple analytes with extremely high sensitivity. The basic principles for the biosensor device are evanescent wave sensing and interferometry based on Mach-Zehnder Interferometers (MZIs). For a MZI with a sensing length of 1000μm the theoretical sensitivity is 3.1 × 10-7 Refractive Index Units (RIUs). Sensitivity is further increased by incorporating spiral waveguides to increase the length of the interferometer arms within a given wafer footprint. A spiral of four turns gives a sensing arm length of 3145μm, which takes up an area of 0.06mm2 and gives a theoretical MZI sensitivity of 9.9 × 10-8 RIUs. The sensitivity of detection of the biosensors developed is at least 10-100 times more sensitive than that of current commercial products. Parallel detection is achieved by exciting multiple sensors in parallel using a 1xN Multimode Interferometer (MMI), where N≤12. Y-junction splitters and Directional Couplers (DCs) are evaluated for splitting purposes. A spot-size converter or grating coupler is used to input and output light to and from the optical waveguides.
2. Keywords: Directional
Coupler; Lab-on-a-Chip; Mach-Zehnder Interferometer; Multimode Interferometer; Spot
Size Converter
1. Introduction
Optical biosensors can be defined as sensor devices which make use of optical principles for the transduction of a biochemical interaction into a suitable output signal. A substantial number of measurements can be made based on emission, absorption, fluorescence, refractometry and polarimetry. Based on the measurements taken optical biosensors can be conveniently categorized into three types; Surface Plasmon Resonance (SPR), Evanescent Field (EF) and photonic crystal. This paper presents the design and optimization of evanescent field sensor based on Mach-Zehnder interferometers. To enable parallel detection and identification of multiple analytes in multiple samples, a splitter in the form of a Multimode Interferometer (MMI), Y-junction splitter or Directional Coupler (DC) is used where each output is connected to a sensor. To couple light in and out of the waveguides, a spot size converter or grating coupler is used. (Figure 1) shows the schematic layout of the proposed biosensor which is constructed with an input coupler, a 1×N splitter, spacers, sensors placed within a microfluidic channel, and output couplers.
(Figure 2) illustrates the various components of an optical biosensor mentioned in the literature and which are reviewed and optimized in this paper. Each component was simulated and optimized for maximum sensitivity and minimum loss for the fundamental vector TM-like mode. However, the structure with the best simulation result is not always the easiest or cheapest to fabricate. Therefore, considering all advantages and disadvantages each component was carefully selected for the final optical biosensor layout.
2. Theory
Light is guided in a high core-cladding refractive index optical waveguide by total internal reflection. During this some of the light travels in the cladding of the waveguide with an amplitude that decays exponentially outwards from the core. The field that penetrates into the upper cladding is the Evanescent Field (EF) that is the basis on which the device operates as a biosensor. This is illustrated schematically in (Figure 3) for the TE polarization. In Figure 3a, the evanescent field is exposed to the upper cladding which can be a solution of refractive index n1 and this type of sensing is referred to as homogenous sensing or bulk sensing. In Figure 3b the evanescent field is exposed to a thin film, also known as an adsorbed layer, attached to the surface of the waveguide with a refractive index n4. This thin film can be composed of antibodies and/or chemical linkers that attach the antibody to the surface of the waveguide. This type of sensing is referred to heterogeneous sensing or surface sensing and is used when the target molecules bind to the antibodies by a chemical reaction. Figure 4 shows the proposed side view of a sensing channel in an evanescent field biosensor.
At the sensing window of the sensor, the waveguide surface is functionalized with antibodies that are specific to the binding of the target molecule. When an analyte is inserted to the cladding at the sensor window, target molecules bind with the receptor molecules changing the refractive index on the surface of the waveguide in the cladding. This results in a change in the effective index of the waveguide (ΔNeff), causing a phase change in the light after propagating for a length L, due to the presence of the evanescent field. Figure 4 also illustrates a microfluidic channel, where simple microfluidic channels and reservoirs were successfully made of SU-8 by photolithography and sealed with Polydimethylsiloxane (PDMS). Silicon based waveguides have been widely researched for biocompatibility and biomedical applications [1,2,3]. Meanwhile SOI material has well established standard fabrication procedures [4,5]. Therefore, SOI material was chosen for the proposed optical biosensor.
3. Simulation Results
3.1. Optimized Core Height of a SOI Waveguide
The first stage of the design was to optimize the SOI waveguide core height to achieve maximum sensitivity. The core height was varied for ridge waveguides of widths 0.5μm and 1μm as illustrated in Figure 5. A width of 0.5μm was used as the waveguide is single mode at the proposed operating wavelength of 1.55μm for this width [6], while a laterally multi-mode ridge width of 1μm waveguide structure was used for comparison. A 2μm oxide (SiO2) thickness was used for this project due to wafer availability. The standard industrial oxide thickness is between 1-3μm [5] and in sub-section 3.5 it is shown that an oxide thickness in this range presents low substrate leakage at an operating wavelength around 1.55μm, resulting in an acceptably low propagation loss.
The core height was varied from 0.14-0.34μm to ensure the waveguide has vertically single mode operation. The change in cover index, Δn1, is 1.32-1.33 which assumes the solvent is water based [7]. The sensitivity was calculated using Equation 1 for the vector TE-like and TM-like modes at an operating wavelength of 1.55μm using a commercial software FIMMWAVE. The results are presented in Figure 6.
From Figure 6, it is shown that the sensitivity obtained for the vector TM-like mode is much higher than that obtained for the vector TE-like mode. The vector TM-like mode sensitivity obtained for waveguide widths of 0.5 μm and 1 μm are similar, while the highest sensitivity is obtained when the core height is in the range of 0.18-0.22 μm. A core height of 0.22μm is preferred in order to take advantage of standard industrial SOI fabrication processes [5].
3.2. Optimization of Core Width
In the previous sub-section, the optimum core height was selected to be 0.22 μm. Now the core width was varied within 0.16-0.60μm for the vector TE-like mode and varied within 0.25-0.60 μm for the vector TM-like mode to ensure single mode operation of the waveguide for each polarization. The sensitivity was calculated as described in the previous sub-section and results are presented in Figure 7a. In Figure 7b, when the width of the waveguide is varied the sensitivity of the waveguide changes significantly for the vector TE-like mode whereas the sensitivity is not affected considerably for the vector TM-like mode. Keeping ease, cost and tolerance of fabrication in mind the waveguides in this research will have a fixed core height of 0.22μm and a width of 0.5μm and will operate in the vector TM-like mode with a high waveguide sensitivity of ≈0.5.
3.3. Heterogeneous Sensitivity
As explained in section 2, homogeneous sensing occurs when the entire upper cladding (n1) index changes, causing an effective index change (ΔNeff) in the waveguide. In an immunosensor the waveguides in the sensing window are covered in antibodies which bind to receptor molecules changing the refractive index at the surface of the waveguide. The region that undergoes a change in refractive index is known as the adsorbed layer or adlayer and is illustrated in Figure 8a. The thickness, t of the adsorbed layer was varied from 0 to 1μm and the sensitivity at an operating wavelength of 1.55μm was calculated using Equation 1. The results are illustrated in Figure 8b. From Figure 8b it is shown that the sensitivity has no change when the adsorbed layer thickness is zero. As the thickness increases the sensitivity increases until it reaches its maximum sensitivity of 0.149 and 0.5 for the fundamental vector TE-like and TM-like modes respectively. The thickness of the adsorbed layer at maximum sensitivity is 0.4μm and 0.7μm respectively which means after this point the waveguide would effectively behave like a homogenous sensor.
3.4. Slot Waveguides Vs Ridge Waveguides
Slot waveguides, as illustrated schematically in Figure 9, have been mainly used in microring resonators for biosensing applications [8,9]. More recently they have also been explored as planar slot waveguides for fluorescence or absorption based optical sensing [10,11,12]. Dell’Olio and Passaro [11] performed a detailed analysis of this. In this sub-section slot waveguide sensitivity with an edge to edge gap, g, of 100nm is compared to ridge waveguide sensitivity for the vector TE-like and TM-like modes at an operating wavelength of 1.55 μm. The sensitivity was calculated for varying core widths for a change in cover index, Δn1, of 1.32-1.33. The results are illustrated in Figure 9.
The TM-like mode supported by the Si-wire is more sensitive than the TE-like mode when the width w>0.31μm and the TM-like mode supported by the slot waveguide is more sensitive than the TE-like mode when w>0.33μm. The slot waveguide is more sensitive than the Si-wire for the TE-like mode but less sensitive for the TM-like mode. To achieve the highest sensitivity of 0.77 the device would have to be designed with slot waveguides with g=100nm and w=0.23μm and operate in the TE-like mode. However, the slot waveguides must be integrated into the ridge waveguide system to make use of their high sensitivity in the TE-like mode. The problem of coupling light into a slot waveguide is addressed in the literature [13,14,15,16] by designing efficient strip-slot mode couplers/converters for SOI devices.
The optimum strip and slot waveguide height and width for this research would be 220 × 500 nm and 220 × 230 nm respectively for the fundamental TE-like mode at an operating wavelength of 1.55μm. In theory a strip-slot coupler can be designed with near 100% efficiency which would allow an increase in sensitivity by 48% for the fundamental TE-like mode, but experimentally there would be some loss caused by the strip-slot coupler. Incorporating slot waveguides would not only require an additional strip-to-slot coupler but also polarization sensitive components such as the MMI coupler would have to be redesigned and optimized for the TE-like mode. Long and narrow waveguides slot waveguides would require more sophisticated and expensive fabrication techniques. Therefore, considering mass production and high reproducibility, working with silicon wire waveguides is preferable.
3.5. Propagation Loss
As light propagates through the waveguide some loss would be caused even though SOI wires have good confinement. This loss is called the propagation loss and is caused by substrate leakage, scattering at sidewall roughness and material absorption. Pieter Dumon et al. [17], showed the substrate leakage of a SOI Si-wire as a function of buffer thickness (silicon dioxide) for varying core widths for the vector TE-like mode only. In this sub-section, the substrate leakage for an SOI ridge waveguide of core height and width of 0.22 μm and 0.5 μm respectively is calculated for varying SiO2 buffer thickness for both vector TE-like and TM-like modes at an operating wavelength of 1.55 μm. The results are illustrated in Figure 10. The vector TE-like mode substrate leakage is similar to that presented by Pieter Dumon et al. [17]. It can be observed that the substrate leakage is larger for the vector TM-like mode and for both modes the loss decreases with buffer thickness.
3.6. Curvature Loss
As shown in the layout of device in Figure 2, the proposed biosensor consists of S-function waveguides, Y-junction splitters, Directional Couplers (DCs), Mach-Zehnder Interferometers with spiral waveguides, all of which require curved or bent waveguides with the smallest possible radius of curvature to achieve a compact device. Light propagating in a bend will be subject to radiation loss [18,19,20] and therefore it is important to determine bending loss as a function of bend radius and hence define the minimum allowable bending radius.
In the literature the minimum bend radius for an SOI waveguide operating at 1.55μm with a fundamental vector TE-like mode was reported to be 2μm, where the loss was 0.15dB for a core of 400×200nm core and 0.6dB for a 300nm square core, while there was no bend related loss when the radius was 5μm or more [15]. Yurii Vlasov et al. [19] showed that when R=5μm, the bend loss is ≈0dB/turn at a wavelength of 1.55μm which is similar to the 0.02dB/cm shown by Zhen Sheng et al. [20] for the vector TE-like mode. Both Vlasov and MacNab [19] and Zhen Sheng et al. [20] did not record the losses for the vector TM-like mode for a wavelength greater than 1450nm as the loss increases exponentially for such narrow radii, as small as 5μm. Figures 11(a, b) show the bending loss of a Si-wire as a function of bend radius for the fundamental vector TE and TM-like modes respectively. Results were obtained from the FIMMWAVE 3D mode solver for the cases when the Si-wire is exposed to air and solution.
From Figure 11(a, b), for the vector TE-like mode the bending loss increases as the index of the cover increases from 1 to 1.33 whereas the opposite is true for the vector TM-like mode. For both vector TE and TM-like modes the bending loss decreases exponentially as the bend radius increases. For the quasi-TE mode and R=5μm, the bending loss is 0.024dB/cm and 0.142dB/cm for nc=1 and nc=1.33 respectively which is well below the straight waveguide propagation losses discussed in sub-section 3.5. The curved waveguides in this research will have a cover index of nc=1.33 or greater and from Figure 11b for bend radius greater than 11μm the loss is negligibly small when nc=1.33. Therefore, for the vector TM-like mode operation, the minimum bend radius was fixed at 15μm.
3.7. Multimode Interference (MMI) Couplers
Multimode Interference (MMI) devices are based on the self-imaging principle where the input image is reproduced into multiple images at the output with high uniformity [21,22]. MMI devices have wide optical bandwidths (in the context of telecommunication systems), low cross talk and are widely used in integrated optics as power splitters [23,24], modulators [25], multiplexers [26], switches [27] and in Mach-Zehnder Interferometers [24]. In this research, parallel sensing of biological fluid is achieved by connecting Mach Zehnder interferometers to the outputs of an MMI coupler. Figure 12 shows a schematic illustration of a 1x2 MMI coupler.
Inside the MMI cavity the beam is split into numerous modes and the symmetric intensity pattern is repeated periodically at the intervals of Λ along the guide. The approximate MMI length, LMMI is calculated by Equation 2 [22]:
where N is the number of outputs, neff
is the effective refractive index of the MMI cavity, WMMI is the
width of the MMI cavity and λ0 is the operational wavelength. The N
images formed are equally spaced across the multimode guide. The transverse
distance between the images is known as the ‘Pitch’,
where Pin is the input power and Pn is the power of the nth port and Pmin and Pmax are the maximum and minimum power observed in the N ports. Insertion loss and non-uniformity calculated for (1×N) MMI couplers up to 1×16 for the vector TM-like mode are presented in Table 1. Table 1 shows a linear increase in LMMI and Insertion Loss as the number of outputs, N, increases. The non-uniformity increases as N increases except for N=8, where it is slightly higher than expected.
3.8. Tapered MMI Coupler
In the previous sub-section, it was shown that for a 1x2 MMI coupler with input and output waveguide width of 0.5 μm the insertion loss is 1.35 dB. Table 2 shows the insertion loss of the MMI coupler decreases as the input and output waveguide width, w increases. The non-uniformity remains at 0 dB. From Table 2 it can be concluded that a wider w gives a lower insertion loss. However, in sub-section 3.2 the width w was optimized at 0.5 μm for sensing. Therefore, a tapered input and output waveguide is studied as illustrated in Figure 13.
The waveguide width, w and taper width and, Tw are fixed at 0.5 μm and 2μm respectively. The LMMI is optimized for and input/output waveguide width of 2 μm and a linear taper of optimum length, TL can be designed separately and then combined to save computational time. Thus, a linear taper length, TL, of 50μm is considered where the taper length is long enough to be adiabatic and the insertion loss caused by the taper is assumed negligible. Table 3 shows the insertion loss and non-uniformity for MMI couplers of up to 12 outputs where the MMI input and output waveguide width, TW was fixed at 2 μm. From Table 3 it can be seen that both the insertion loss and non-uniformity increase as N increases, but these values are much lower compared to the corresponding values presented in Table 1 where w was fixed at 0.5μm.
3.9. S-Function Waveguides
As shown in Figure 1, bent waveguides are
used to space the MZI sensors apart and in the construction of the MZI itself.
S-function waveguides are also used in the construction of Y-junction splitters
and Directional Couplers (DCs). In FIMMWAVE the path of the path of the S-function
curve is defined by Equation 5 [32]:
As shown in Figure 14 the software lets the
user input Plhs and Prhs as the left-hand side and
right-hand side offsets from the z axis. Then the user can input the length, L
of the waveguide structure. However, from Equation 5 it is not possible to
directly calculate the radius of curvature and therefore Equation 6 [33] was
used to calculate the radius of curvature at each position along of the curve.
Equations 6-8 can be used to find the minimum radius of curvature, Rmin which is expressed by Equation 9:
This is useful as when Plhs, Prhs
and L become larger the simulation window becomes larger making it time
consuming, and sometimes impossible, to simulate the transmission loss in the
bend using the software. Therefore, using equation 9 the minimum length, Lmin
can be calculated to give
3.10. Y-Junction Splitters
Y-junction splitters are described by two S-function waveguides as shown in Figure 15. Y-junction splitters are used in the construction of the MZI and can themselves be used as a 1xN splitter if cascaded. Table 4 shows the insertion loss and non-uniformity of Y-junction splitters when the outputs are spaced 5, 10 and 15 μm apart for the vector TM-like mode at an operating wavelength of 1.55μm for a fixed Rmin of 15μm. The length and the insertion loss increases as the distance between the outputs, ρ is increased.
3.11. Directional Couplers
Directional Couplers (DCs) are constructed from two propagating waveguides set close enough together that energy passing through one is coupled to the other [34]. DCs can be used to replace Y-junctions in the construction of a MZI and cascaded DCs can be used as a 1xN splitter. In this sub-section, DCs are designed and compared with the equivalent Y-junction splitters in terms of insertion loss and length. Figure 16 illustrates a DC where half the light from the propagating lower waveguide is coupled to the non-excited upper waveguide were d is the edge to edge gap between the two waveguides. In order to compare the Y-junction to the DC, S-function waveguides were added to the DC with an output waveguide separation of ρ.
DCs are sensitive to wavelength and
polarization as the coupling length, Lc for complete power transfer
is expressed by Equation 10 [35].
where ne and no are the
effective refractive indices of the even and odd supermodes supported by the
structure and ko is the free space wavevector.
From Figure 17a, the length increases logarithmically as the edge to edge coupling distance, d, between the coupling and isolated waveguide is increased. Therefore, to obtain a shorter device a smaller d is preferred. From Figure 17b the insertion loss reduces to ≈0 dB as d ≥ 0.4μm. Therefore, d was fixed to 0.45μm and the coupling length was fixed at 12.46μm for the results shown in Table 5. The results shown in Table 5 include S-function waveguides added to physically separate the outputs of the DC. The length and insertion loss both increase as the final waveguide separation ρ increases.
3.12. Mach-Zehnder Interferometer
Evanescent field sensors based on effective refractive index changes generally operate in interference based systems such as Mach-Zehnder Interferometers (MZIs) [36], Young interferometers [37,38] and ring resonators [39]. These devices use different techniques of superposition to extract phase information about the propagating waves. Figure 18 provides a schematic illustration of a MZI.
The incident light is divided into two equal
parts that propagate to the sensing arm and the reference branch respectively.
A change in the effective index of the waveguide (ΔNeff) at the
sensing window, caused by the binding of the target molecule to the antibody,
gives a phase change Δφ in the interaction length, L, as expressed in Equation
11.
where k0 is the free space wave number or the phase constant (2π/λ0).
The reference branch is used to reduce
ambiguity due to external factors such as input power fluctuations, temperature
changes in the environment and non-specific adsorption. Non-specific adsorption
is when other molecules that are not the target molecule bind with the antibody
and cause a phase change. Thus, this reference arm should also be functionalized
with the same antibody as the sensing arm, and a similar analyte but without
the target molecule should flow over it. When the light from the sensing and
reference arms recombines, constructive or destructive interference occurs;
this in turn modulates the light intensity at the single mode output. The
output intensity is related to the phase change
where I0 is the input intensity.
The cosinusoidal variation of the interference pattern can be directly related
to the concentration of the analyte to be measured via the change in effective
refractive index as described in Equation 1. For L=1000μm,
3.13. MZI with Spirals
As presented in Equation 11 the induced phase change increases as the sensing length increases but in the configuration shown in Figure 18, as the length increases the overall length of the device increases. To counter this problem spiral waveguide are used. Spiral waveguides are increasingly gaining popularity and have recently been used for many applications such as spectrometers [41,42], ring resonators [43,44], filters [43,45], chirpers [46] and Bragg gratings [47]. The most common type of spiral used is the Archimedean spiral which can be expressed in parametric form as Equations 13-14.
Figure 19 shows a MATLAB plot the two arms of a MZI designed with Archimedean spirals with Rmin=15μm.
The two arms are spaced apart, the spacing being described using vertical and horizontal offsets. For the design in Figure 19 the vertical offset was fixed at 25μm so that the Y-junction designed in sub-section 3.10 can be used to connect the two arms to create a MZI. The minimum horizontal offset is set at 4 times the maximum R which in this case is 163μm. Therefore, the horizontal offset was fixed at 175μm. The length of one spiral is given by Equation 15 [48].
Table 6 presents the maximum radius, minimum horizontal offset, horizontal offset used, area of MZI, the length of each arm and sensitivity of spiral MZI for 4, 10, 25 and 40 turns. From Table 6, it can be seen that the footprint or area of the MZI is small compared to the length of each arm achieved thus enabling the device to be as compact as possible while achieving the highest possible sensitivity. However, as the length of the arm increases the device becomes more vulnerable to fabrication errors. The increase in sensitivity (RIU) must be weighed against the complexity and yield of fabrication.
3.14. Spot-Size Converter
In this sub-section, a spot size converter is designed to propagate a 2±0.5um diameter light beam at a wavelength of 1.55μm into a 0.5 × 0.22μm cross section SOI waveguide with the help of an SU-8 taper as shown in Figure 20. The 2μm thick SU-8 taper sits on an inverted silicon waveguide taper.
Pu et al. [49] show that a smaller silicon taper tip width reduces loss; however, it is difficult to fabricate extremely narrow waveguides by etching. Therefore, the silicon taper tip width was limited to 100nm in this project. The length of the SU-8 taper is assumed to be twice the length of the inverted silicon taper for simulation convenience at this point. The width and length of the SU-8 taper were varied as shown in Figures 21(a,b) to obtain maximum transmission. Calculations were performed using FIMMPROP for the vector TM-like mode at an operating wavelength of 1.55μm.
It can be seen from Figures 21(a, b), that in order to obtain a 97.7% transmission the taper length must be at least 400nm and the input taper with width of 10μm. Thus, one might set the taper length to 600μm to allow tolerance when cleaving. In the literature spot size converters have been designed with a rectangular shaped polymer on top of an inverted tapered silicon waveguide, the polymer structure is not tapered. Table 7 compares the results obtained above with the dimensions, insertion loss and experimental coupling efficiency results taken from the literature.
From the literature, the lowest insertion loss for the TM-like mode was measured to be 0.36 dB [49]. By tapering the SU-8 polymer, the insertion loss can be reduced to 0.1dB and the coupling efficiency can be increased to 97.7%.
3.15. Grating Couplers
This method is gaining popularity rapidly as aligning the input beam to the waveguide is easier compared to the previous method, even though the input must be at a certain angle as shown in Figure 22. Moreover, the grating coupler is compact.
In order to couple light into the waveguide
the phase matching condition must be met. This means the propagation constant
in the z direction must be the same. This is given by the Equation 16:
Since
One-Dimensional (1D) grating couplers are polarization sensitive and operate in the vector TE-like mode [54,55]. To operate in the vector TM-like mode, Two-Dimensional (2D) gratings are used [54,56]. Once the grating patterns are optimized the coupling efficiency can be considerably improved further by focusing the gratings, apodization, or focused apodization. Frederik Van Laere et al. [57] designed and fabricated a focused grating coupler for the vector TE-like mode with a theoretical efficiency of 37% for an operating wavelength of 1.55μm. Robert Halir et al. [58] designed an apodized grating coupler and Zhenzhou Chen et al. [59] designed a focused apodized grating coupler where both grating couplers were designed for the vector TM-like mode with a coupling loss of 3.7dB and 3dB respectively. Grating couplers are inefficient compared to spot size converters due to mode mismatch and they require an additional etch step; but they allow for compact and high-density integration of photonic devices which is important for the commercialization of the sensor.
4. Conclusion
This paper has successfully presented the simulation evaluation of a label-free Silicon-On-Insulator (SOI) optical biosensor. The initial stage of this work was to investigate the ideal material for the transducer and the type of waveguide to obtain a high sensitivity (ΔNeff/Δnc) of detection while also considering biocompatibility, easy integration, and established fabrication procedures for mass production. Calculations showed that High Index Contrast (HIC) materials such as SOI gave the highest sensitivity of 0.47, when the silicon core height was 0.18μm and the upper cladding index, nc was 1.33 for the vector TM-like mode, at an operating wavelength of 1.55μm. SOI based waveguides are also widely used in biomedical applications [1,2,3], allow easy integration and have well established fabrication procedures [4,5].
Simulation investigation performed using the software FIMMWAVE showed that obtaining a high sensitivity of 0.68 is possible if the silicon ridge waveguide height and width was fixed to 0.22μm and 0.27μm respectively and operated in the vector TE-like mode at a wavelength of 1.55μm. The sensitivity can be increased further to 0.77 by using slot waveguides of height and width of 0.22μm and 0.23μm respectively for the vector TE-like mode at a wavelength of 1.55μm. However, the fabrication of such narrow and long waveguides would require sophisticated and expensive fabrication technologies. For wider waveguides, the vector TM-like mode becomes more sensitive than the TE-like mode. To take advantage of the standard industrial SOI fabrication procedure [5], the silicon ridge waveguide height and width need to be 0.22μm and 0.5μm respectively and operate in the vector TM-like mode at a wavelength of 1.55μm. Due to the availability of SOI wafers two other ridge waveguide dimensions were also used where the silicon ridge waveguide heights were 0.32μm and 0.25μm and the widths were 4μm and 0.8μm respectively.
Propagation loss due to substrate leakage was calculated and showed that as the silicon dioxide thickness separating the silicon guiding layer from the silicon substrate increased, the substrate leakage loss decreased for both the vector TE-like and TM-like modes. The SOI wafers used in this project had a 2μm thick silicon dioxide buffer layer which theoretically gives a substrate leakage loss of 0.23dB/cm for the vector TM-like mode at an operating wavelength of 1.55μm. The curvature loss for the vector TE-like mode increases as the cover index, nc, increases whereas the opposite is true for the vector TM-like mode. For an SOI ridge waveguide of height and width of 0.22μm and 0.5μm respectively and nc=1.33 operating in the vector TM-like mode at a wavelength of 1.55μm, the calculated curvature loss is 0dB/cm (below calculation error) for a radius of curvature greater than 11μm. Therefore, the minimum radius of curvature was fixed at 15μm for all devices in this project.
Splitting light into two or more waveguides was achieved by Y-junction splitters, Directional Couplers (DCs) and Multimode Interference (MMI) couplers. Simulations show that Y-junction splitters require the shortest device length whereas DCs have the smallest insertion loss. Y-junctions splitters are more vulnerable to fabrication errors due to the sharp splitting of waveguides at the junction. To achieve the splitting of more than two outputs the Y-junction splitters and DCs must be cascaded and with each cascaded section the probability of fabrication errors increases. Therefore, each splitting method has its advantages and disadvantages and all three devices were designed, fabricated and tested successfully. Mach-Zehnder Interferometers (MZIs) constructed by Y-junction splitters, straight waveguides and spiral waveguides were designed, fabricated and tested successfully. The longer the MZI the more sensitive the sensor became. Simulations showed that if the adsorbed layer of a heterogeneous sensor is greater than 0.7μm the sensor acts like a homogeneous sensor. For a homogenous sensing MZI with length, L, of 1000μm and ΔNeff/Δnc =0.5 the theoretical MZI sensitivity was 3.1 × 10-7 RIUs. For a MZI with spirals of 10 turns and L=3145μm the theoretical MZI sensitivity was 9.9 × 10-8.
Increasing the sensitivity is not always optimal as the output power intensity is a co-sinusoidal graph which could result in fringe order ambiguity, directional ambiguity and sensitivity fading. This could be avoided by reducing the length of the MZI, thereby reducing the sensitivity, so that the intensity is always located on one arm of the fringe in an intensity graph. Butt coupling and grating coupling methods were used to couple light into and out of waveguides in this project. For butt coupling, a tapered spot size converter was successfully fabricated on top of an inverted silicon taper to efficiently couple the light in and out of the waveguides. Simulations show a 97.7% coupling efficiency and 0.1dB insertion loss for an SU-8 spot size converter tapered from 10μm to 0.5μm with a height of 2μm and a silicon waveguide inversely tapered from 0.1μm to 0.5μm for the vector TM-like mode at an operating wavelength of 1.55μm. The total length of the taper was 600μm.
The grating coupler
used was 35μm in length and 24μm in width. The theoretical and experimental
coupling efficiency was about 4dB and 7.25dB respectively. Grating couplers
require a small area compared to the butt coupling method but have a much
higher insertion loss. Label-free SOI optical biosensors based on the evanescent
wave principle have been previously demonstrated [36,60]. The novelty of the
biosensors in this work lies on the transducer layout, tapered spot size
converters and grating couplers, parallel sensing of up to 20 sensors,
increased sensitivity and the surface functionalization of the silicon
waveguides (not presented in this paper).
Figure 1: Schematic diagram of the proposed biosensor layout which
is made of an input coupler, a 1xN splitter (where N is the number of outputs),
separators, sensors, output couplers and a microfluidics layer.
Figure 2: Possible structures considered for the construction of
an evanescent wave optical biosensor.
Figures 3(a,b): TE Evanescent Field (EF) in a slab waveguide: a) Homogenous or bulk EF sensor and b) Surface or adsorbed molecule layer
EF sensor.
Figure 4: Schematic diagram showing the side view of a sensor
window with target molecules that bind with antibodies coated on the waveguide
causing a change in refractive index in the evanescent field region of the
waveguide.
Figure 5(a,
b): Ridge waveguide with width a) 0.5μm and b) 1μm for varying core heights.
Figure 6: Silicon ridge
waveguide sensitivity for vector and semi-vector TE-like and TM-like modes for
varying core heights.
Figures 7(a,
b): Ridge waveguide a) structure and parameters and b) sensitivity for Vector TE-like and Vector TM-like fundamental
modes for varying waveguide widths.
Figures
8(a, b): a)
Ridge waveguide with adsorbed layer of thickness, t and b) Sensitivity as a function of adsorbed layer thickness for vector
TE-like and TM-like modes for a silicon waveguide of height 0.22μm, width 0.5μm
and an operating wavelength of 1.55μm.
Figure 9: SOI slot waveguide a)
structure and parameters and b)
sensitivity as a function of width for quasi-TE and quasi-TM simulated with
FIMMWAVE at a wavelength of 1.55μm.
Figure 10: Substrate leakage loss as a function of buffer (SiO2)
thickness for the fundamental vector TE- like and TM-like modes at an operating
wavelength of 1.55μm.
Figures 11(a, b): a) Vector-TE mode and b)
vector-TM like bending loss as a function of bend radius, R for when a
500×220nm waveguide is exposed to air and solution of refractive index of 1.33
at an operating wavelength of 1.55μm.
Figure 12: Schematic diagram of a 1x2 MMI structure.
Figure 13: Illustration of a 1x2 MMI with tapered input and output
waveguides.
Figure 14:
FIMMPROP CAD layout of a S-function
waveguide.
Figure 15: Y-junction splitters.
Figure 16: DC coupler designed
for the vector TM-like mode where d=0.45μm, coupling length=11.4μm, ρ=5μm,
length of S-function waveguides=13μm and the operating wavelength=1.55μm.
Figures 17(a,
b): Vector TM-like mode (Ey) results for DC a) coupling length as d is varied and b) insertion loss as gap is varied for
an operating wavelength of 1.55μm.
Figure 18: Schematic illustration of a Mach-Zehnder Interferometer
(MZI).
Figure 19: MATLAB layout of the sensing and reference arms of the
MZI with spiral waveguides where each spiral has 4 turns.
Figure 20: Spot size converter with a tapered SU-8 layer on top of
an inverted tapered silicon waveguide.
Figures
21(a, b): Spot size converter transmission (a) function of SU-8 taper length when
taper width=10μm and (b) function of
SU-8 taper width when taper length=600μm for the vector TM-like mode at an
operating wavelength of 1.55μm.
Figure 22: Illustration of grating coupling.
N |
WMMI (μm) |
LMMI (μm) |
Insertion Loss
(dB) |
Non-uniformity
(dB) |
2 |
10 |
68.9 |
1.35 |
0 |
4 |
20 |
134.5 |
1.66 |
0.01 |
8 |
40 |
268 |
2.3 |
1.11 |
12 |
60 |
401.8 |
2.91 |
1.01 |
16 |
80 |
537.4 |
2.97 |
1.7 |
Table 1: Insertion loss and
non-uniformity for single mode waveguide MMI splitters the vector TM-like mode
at a wavelength of 1.55μm.
Insertion
loss (dB) |
Non-uniformity
(dB) |
||
1.35 |
|||
1 |
68.4 |
1 |
0 |
1.5 |
68 |
0.21 |
0 |
0.08 |
0 |
LMMI
(μm) |
Insertion
Loss (dB) |
Non-uniformity
(dB) |
||
2 |
10 |
69.7 |
0.08 |
0 |
4 |
20 |
136.5 |
0.19 |
0.06 |
8 |
40 |
270 |
0.44 |
0.44 |
12 |
60 |
405.6 |
0.57 |
0.12 |
Table 3: Insertion loss and non-uniformity for MMI splitters with
input/output waveguide width of 2μm.
ρ (μm) |
Length (μm) |
Insertion loss
(dB) |
Non-uniformity
(dB) |
5 |
14 |
0.26 |
|
10 |
20 |
0.31 |
0 |
20 |
28 |
0.44 |
0 |
Table 4: Insertion loss and non-uniformity of optimized
Y-junction splitters.
ρ (μm) |
Length (μm) |
Insertion loss (dB) |
Non-uniformity (dB) |
5 |
24 |
0.04 |
0 |
10 |
30 |
0.11 |
0 |
20 |
38 |
0.3 |
0 |
Table 5: Insertion loss and non-uniformity of optimized
Y-junction splitters.
Turns |
Maximum Radius (μm) |
Minimum
horizontal offset (μm) |
Horizontal
offset used (μm) |
Area (Length ×
width) (μm × μm) |
Length of arm
(μm) |
Sensitivity (RIUs) |
4 |
40.84 |
163.3 |
175 |
0.03mm2(338.4 × 81.7) |
1 175 |
2.6 × 10-7 |
10 |
59.69 |
238.7 |
250 |
0.06mm2(488.8 × 119.4) |
3 145 |
9.9 × 10-8 |
25 |
106.81 |
427.3 |
450 |
0.19mm2(877.2× 213.6) |
11 192 |
2.8 × 10-8 |
40 |
153.94 |
615.8 |
625 |
0.38mm2(1,239.2
× 307.8) |
23 655 |
1.3 × 10-8 |
Table 6: Maximum radius, Minimum horizontal offset, horizontal
offset used, area and length of arm of spiral MZI for turns=4, 10, 25 and 40.
Paper |
Silicon taper |
Mode |
Upper cladding |
Insertion Loss
(dB) |
Coupling
efficiency |
|||
Height × width
(nm) |
Silicon tip (nm) |
Length (μm) |
material |
Height × width
(μm) |
||||
[50] |
220 × 450 |
75 |
150 |
TE |
polymer |
2 × 2 |
1 |
- |
[51] |
350 × 300 |
50 |
60 |
TE |
SiO2 |
1.5 × 5 |
- |
66% |
[49] |
250 × 480 |
40 |
300 |
TE |
SiO2+Polymer |
3.4 ×3.4 |
0.66 |
- |
TM |
0.36 |
- |
||||||
340 × - |
40 |
200 |
TE |
SU-8 |
3 × 5 |
- |
60% |
|
TM |
- |
60% |
||||||
This work |
220 × 500 |
100 |
600 |
TM |
SU-8 |
2 × tapered from
10 to 0.5 |
0.1 |
97.70% |
Table 7: Experimental results for spot size converters from literature and theoretical results from this project.
1.
Coffer JL (2014) Semiconducting silicon nanowires for
biomedical applications. Elsevier Cambridge.
12.
Barrios, CA (2009) Optical slot-waveguide based biochemical
sensors. Sensors 9: 4751-4765.
32.
Design P (1997-2012)
FIMMPROP Version 5.4.1 Manual.
1.
(2017) Arc Length of a Curve.
53.
Reed GT, Knights AP (2004) Silicon photonics: An
introduction. John Wiley & Sons, West Sussex, UK.