Sensitivity and Limit of detection of biosensors based on ring resonators

In this article, we compared different silicon oxynitride (SiON) microring-based photonic structures for biosensing applications. We varied the refractive index of SiON as well as the dimensions of the resonators, and analyzed their volumetric sensitivity and limit of detection by using glucose-water solutions. By functionalizing the surface of these sensors, we also investigated the specific detection of Aflatoxin M1. We observed binding of aflatoxin at concentrations as low as 12.5nM, and regeneration of a single functionalized sample up to 17 times by using glycine solution. The results open up the path for designing cost effective biosensors for a fast and reliable sensitive analysis of AFM1 in milk.


Introduction
Silicon photonic biosensors are ideal candidate for the development of integrated electro-optic microfluidic chips (Lab on a chip, LOC) devices [1,2]. Their numerous advantagessuch as miniaturization, robustness and accuracycan lead to devices able to perform completely automated biological analysis and clinical diagnostics for pharmaceutical and biotechnology companies. Nowadays, the most widely used optical biosensors are based on Surface Plasmon Resonance (SPR) techniques [3], but nanophotonic devices like ring resonators and optical waveguide gratings are in full expansion [4].
One application in which a silicon photonic biosensor could be a breakthrough innovation is the sensing of toxins in the dairy production. The European Community regulation, in fact, limits the maximum allowable concentration of Aflatoxin in milk products to 50 ng/kg. Presently, the screening procedure involves Enzyme-Linked ImmunoSorbent Assay (ELISA) tests [5], and the suspicious samples need further investigations with High-Performance Liquid Chromatography (HPLC) tests [6], which are costly and time-consuming processes.
Within the European project Symphony, we are developing an optical biosensor to detect Aflatoxin in milk with a fast, cheap and reliable procedure. Since the presence of casein micelles and milk fat globules could affect the performances of our sensor, a pre-purification module is being developed in order to deliver to the photonic sensor a solution cleaned by proteins and contaminants [7]. Therefore, in the following, we will characterize the sensor with model solutions.
In this work, we designed and measured microring-based photonic biosensors. These devices have silicon oxynitride (SiON) as waveguide core material, in order to work in the visible range, and are fabricated by standard CMOS processing. We analyzed their Sensitivity (S) and Limit of Detection (LOD) as a function of the waveguide composition and dimensions. Furthermore, we performed sensing measurements on samples functionalized with biorecognition agents, DNA-aptamers, specific for the detection of Aflatoxin M1 [8,9]. We performed measurements with Aflatoxin M1 in a buffer solution on samples with and without aptamers, in order to observe the binding of Aflatoxin on the surface of the sensor. We repeated and confirmed these measurements regenerating the samples using Glycine solution.

Fabrication process
For the microring fabrication, SiON films were deposited by plasmaenhanced chemical vapor deposition (PECVD) on five 6-in. 625 μm thick crystalline Si wafers with a 4 μm thick buffer oxide layer. For one wafer, the SiON refractive index was chosen to be 1.8, while, for the others, the SiON refractive index was 1.66. Table 1 reports the thicknesses and refractive indices of all wafers.
The waveguides, ring resonators and directional couplers structures were defined using standard UV-lithography and reactive-ion etching techniques. An annealing step of 1.5 h at 1050°C was also performed in order to remove hydrogen bonds from the material. Such process should also decrease the losses of the structures. Finally, the processed wafers were covered with a 1 μm TEOS film, used as a cladding layer.
In order to create the sensors, we opened the cladding layer around the resonators through a BHF wet etching. This allows the functionalization of the rings as well as the sensing measurements through their exposure to the ambient and to the molecules to be analyzed.

Sensor design
A scheme of the photonic sensor architecture is presented in Fig. 1. The input signal is directed to four different microrings. The width of the bus waveguide was fixed to 900 nm, while the resonator waveguide width to 1000 nm. The use of waveguide directional coupler for the splitters allows using a large gap distance between the waveguides (in our case 600 nm) and avoids the use of high lithography resolution. The same directional couplers are used to couple the bus waveguide to the resonator waveguide.
As a resonator structure, we used racetrack shaped ring resonator. The gap between the bus and the ring waveguides was set to 600 nm, while the curvature radius has been imposed to be R = 100 μm in order to have negligible bending losses.

Functionalization process
The functionalization procedure performed on the sensors is based on wet silanization protocol. After a cleaning process (with a Piranha solution) to remove organic contaminations, the samples were immersed in 0.01% v/v of GPTMS (3-glycidoxypropyl methyldiethoxy silane) in anhydrous toluene at 60°C for 10 min. Then an amino-terminated DNA-aptamer (5′-NH 2 -(CH 2 ) 6 -GT TGG GCA CGT GTT GTC TCT CTG TGT CTC GTG CCC TTC GCT AGG CCC ACA-3′) at 100 μM in phosphate buffer (50 mM, ionic strength 300 mM, pH 8) was incubated on silanized surfaces for 2 h. The aptameric sequence with a k D of 10 nM was identified by NeoVentures Biotechnology Inc. [8]. The aminomodified sequence is HPLC purified and was purchased from IDT Integrated DNA Technologies (Leuven, Belgium). Finally, an ethanolamine passivation at 1 mM for 30 min was applied.

Measurement setup
The sensors were characterized using a single mode tapered fiber placed on a piezoelectric movement as input. For collecting the light at the output of the different resonators simultaneously, we used a fiber array placed on a piezoelectric movement. Glycerol was placed between the sample and the fiber array and acts as index matching gel in order to increase the coupling efficiency between the fibers and the output waveguides. As light source, we used a single mode and polarization VCSEL at 850 nm connected to a single mode fiber. The polarization of the input light was controlled using a 2-paddles polarization controller. For the analysis of the data, we used transimpedance amplified Siphotodetectors and an eight channel USB oscilloscope connected to a computer. A visible camera mounted on top of an optical microscope was also used for the alignment process.
For sensitivity and sensing measurements, we used a homemade PDMS microfluidic flow cell, with a volume chamber of less than 0.5 μL, connected to a VICI M6 liquid handling pump.

Sensitivity and Limit of Detection
The spectral sensitivity of the sensor is defined as S = Δλ/Δn where Δλ represents the shift of the sensor resonance in nm, and Δn the change of the refractive index solution flowing on top of the sensor. In our experiments, we measured the sensitivities of the ring resonators as a function of the SiON refractive index and of the waveguide thickness. The measurement was achieved by exposing the sensors to glucose-water solutions of various concentrations, whose refractive index can be easily estimated [10]. The shift of the resonance of the sensor as a function of the flowing solution allows us to determine the sensitivity of the resonator. The results of these measurements for all wafers are presented in Fig. 2. The highest sensitivity was found in the case of n SiON = 1.8 with S = 112 nm/ RIU. For n SiON = 1.66, the best thickness was found in the case of BS2 wafer, with S = 82 nm/RIU, and with a very good reproducibility for what concerns similar resonators on the same chip (differences in the order of 1%).
To link these results with the Limit of Detection (LOD) of our sensors, we measured the noise level of our experimental setup. The LOD is defined as the minimum input quantity that can be distinguished with more than 99% fidelity, and can be calculated as LOD = 3∈/S, where ∈ is the output uncertainty. Using this formula, we measured a LOD of 3 × 10 −6 RIU in the case of n SiON = 1.66, and 1.6 × 10 −6 RIU for n SiON = 1.8. According to these results, we decided to perform sensing measurements on BS2 sample, which has the best sensitivity concerning samples with n SiON = 1.66.

Aflatoxin sensing measurements
To perform Aflatoxin-sensing measurements on BS2 samples, we initially filled the microfluidic chamber with a buffer solution. In our case, the buffer solution is mainly based on 2-(Nmorpholino)ethanesulfonic acid (MES) with a certain amount of dimethyl sulfoxide (DMSO) that depends on the Aflatoxin. We then injected the solution containing Aflatoxin AFM1 at a known concentration, and monitored the drift in time of the resonance due to the capture of Aflatoxin from the functionalized biosensor. Finally, we injected again the buffer solution. All measurements were done with a flow of 3 μL/min.
Two solutions were investigated, one with AMF1 and one without, in order to see the resonance shift due to the binding of Aflatoxin on the surface of the chip.
It is important to note that after each Aflatoxin injection, two or three Glycine injections were effectuated, in order to detach the toxin from the sensor and regenerate the aptamers. The regeneration performances of Glycine concerning this aptamer were already demonstrated in [11]. An example of such measurement for Aflatoxin concentration of 12.5 nM (DMSO concentration 0.1% v/v) is presented in Fig. 3. For what concerns the shift of the resonance, we can observe a clear difference between the two solutions. In the case of the solution with Aflatoxin, the drift is more pronounced in the first minutes of the injection, and then slowly decreases until reaching, in the last minutes, a slope similar to the one of the sample without toxin. Such behavior can indicate a binding of the Aflatoxin to the aptamer, as the adsorption of surfaceactive molecules at the interface of a two-phase system can be modeled with the classical Langmuir isotherm, which describes an exponential binding profile [12].
In order to confirm this, we repeated several times this experiment for all the solutions and, as shown in Fig. 3, the repeated measures overlap perfectly. We were able to reiterate this experiment and observe a binding effect down to a minimum concentration of 1.58 nM.

Conclusion
In this article, we designed and optimized microring-based photonic structures for biosensing applications, and were able to observe binding of Aflatoxin M1 for concentrations as low as 1.58 nM. The high performances of this photonic biosensor (high sensitivity, LOD ∼ 10 −6 RIU, good regeneration) will allow the creation of a fast, inexpensive and reliable sensor for applications in the areas of environmental monitoring, agrofood, clinical medicine and healthcare.