2.1.

Hydrogenated Amorphous Silicon Photonic Sensing Platform

Different kinds of transparent optical materials like nitrides, polymers, glasses, and others have been already used to fabricate photonic resonators for lab-on-chip sensing applications.^8,9 However, there are essential advantages of employing a high refractive index material like silicon, which are of economic and physical nature.

The main economic reason is the compatibility of advanced CMOS-technology with well-developed infrastructure and long-time process experience, which will allow sharing existing facilities with electronics, or to combine electronic and photonic systems on the same substrate. Furthermore, the high index contrast (HIC) between waveguide core and cladding materials (e.g., Si and ${\mathrm{SiO}}_2$ with $\mathrm{\Delta }n\ge 2$) permits a high integration density and thus photonic systems with low footprint so that hundreds of sensor arrays with multiple channels can be cost effectively fabricated on a single wafer. This makes the silicon platform interesting for applications like point-of-care diagnostics. However, the HIC is just as much beneficial in terms of the sensor sensitivity. The strong mode confinement in single-mode photonic wire waveguides with typical dimensions of $500\times 200\mathrm{nm}$ allows for small bending radii $R\le 5\mu \text{m}$ with negligible losses of $\approx 0.1\mathrm{dB}$ per revolution. As a result of that, a large free spectral range of $\approx 18\mathrm{nm}$ can be achieved with such ring resonators, which facilitates a distinct detection of large analyte changes and favors the interrogation of several resonator-based sensors with one readout waveguide.

Addditionally, the tight light guidance implies that there is a strong electric field discontinuity at the low and high refractive index layer interfaces and thus the evanescent field can effectively interact with the analyte medium which covers the photonic sensor. The sensitivity enhancement can be expressed in terms of the evanescent field strength according to

Fig. 1

Results of finite element method (FEM) simulations: Fundamental qTE (a) and qTM (b) guided mode intensity plots for an a-Si:H photonic waveguide with $480\times 200\mathrm{nm}$ dimensions and water cladding with electric field orientations inset as arrows. Normalized electric field strength line plots of the fundamental modes for the width (c) and the height (d) of the photonic wires.

Table 1

Effective mode indices and confinement factors for both fundamental modes of a 480×200  nm dimensioned a-Si:H photonic wire with oxide under and aqueous top cladding.

2.2.

Sensing Principle and Figures of Merit

Photonic evanescent field refractive index sensors can be categorized due to different sensing principles as shown in Fig. 2.¹¹ Surface sensing is a method where selective receptor molecules are immobilized on a functionalized photonic device surface so that only target molecules will selectively stick to the receptors. This way the propagation of the guided mode ($N_{\mathrm{eff}}$) is slightly altered and hence, the response of the photonic sensor is a measure of the adsorpted molecule amount. The surface sensing technique is very promising for lab-on-chip or point-of-care biosensing applications. Compared to that, homogeneous sensing might be employed to measure gases and chemicals. In homogeneous sensing, the photonic sensor is covered with a cladding medium, e.g., water or air, and a change of a specific concentration in the medium (e.g., NaCl) is detected by the refractive index difference $\mathrm{\Delta }n$ of the analyte. For both the methods, the interrogation of the sensor can be performed with spectral transmission measurements and the signal can be evaluated either by a resonance peak shift or by an increase in absorption, for example, induced by gases in the mid-infrared. In this work, homogeneous sensing was employed in order to detect the refractive index change by measuring the resonance wavelength shift due to different NaCl concentrations. The intrinsic sensitivity ($S_{\text{RIU}}$) of the resonator-based refractive index sensor can be defined as

Fig. 2

(a) and (b) Illustration of the sensing principle by a change of cladding refractive index which results in a wavelength shift. Scheme of (c) homogeneous sensing method and (d) of surface sensing method.

From the set of Eqs. (1)–(4), it is evident that a high sensitivity is achieved when there is a strong modal overlap of the guided mode with the sensing medium and that additionally a high $Q$ resonator will enhance the LoD. However, a high $Q$ factor does not necessarily improve the sensitivity, especially for water-based solutions, because water adds substantial propagation loss ($\approx 10{\mathrm{cm}}^{-1}$ at 1550 nm), and a high quality factor often results from a stronger mode confinement in the resonator core and hence less light can effectively interact with the sensing medium.^12,13

Based on these considerations, the homogeneous sensing experiments were carried out with microdisk resonators which exhibit moderate $Q$-factors of ${10}^4$ although we have already fabricated critically coupled disk resonators with higher intrinsic $Q$ of $\ge {10}^5$ for qTE and $\ge 0.9\times {10}^6$ for qTM with amorphous silicon.¹⁴ In order to achieve a strong interaction of the evanescent field with the sensing medium, the following disk resonator experiments were conducted for the qTM-mode. The disk radius was chosen to be $R=10\mu \text{m}$ so that a sufficiently large free spectral range ($\mathrm{FSR}=10\mathrm{nm}$) for resonances of the same modal order was achieved. The coupling gap between disk and waveguide was designed to be 280 nm for weak coupling.

2.3.

Functionalization of the Resonators

The sensor chip was functionalized through following process steps for the detection of BSA proteins. First, the chip surface was cleaned with Piranha-etch followed by several wash steps. Then, the a-Si:H sensor surface was modified by a silanization step with alkoxysilane molecules (APTES). Glutaraldehyde was utilizied as an amine-reactive homobifunctional-crosslink in order to allow the binding of the targeted BSA molecules. The process procedure is schematically summarized in Fig. 3(a). In order to verify the binding of the BSA-proteins on a-Si:H chips, fluorescent measurements of the supernatant were performed. The FluoroProfile Protein Quantification Kit (Sigma-Aldrich, Germany) used is based on the epicocconone. The epicocconone represents a fluorescent probe based on a polyketide skeleton isolated from the fungus Epicoccum nigrum and produces a large increase in fluorescence quantum yield upon binding to detergent-coated proteins in solution.¹⁵ The protein quantification was carried out in accordance with the supplier’s protocol. An atomic force microscope (AFM) picture with a $5\times 5\mu \text{m}$ scan area of a sample with adsorbed BSA proteins is shown in Fig. 3(b).

Fig. 3

(a) Functionalization procedure using silanization (APTES) and amine-reactive homobifunctional-crosslinking. (b) Atomic force microscope picture of immobilized bovine serum albumin (BSA) on a-Si:H sample.

3.1.

Layer Deposition and Photonic Sensor Fabrication

Standard crystalline silicon wafers (10 cm) with a 3 μm thermal grown bottom oxide cladding layer were used as substrates. Thus, for this refractive index contrast and layer thickness, substrate leakage is effectively supressed ($\le 0.01\mathrm{dB}/\mathrm{cm}$) for both fundamental modes (qTE and qTM). The PECVD chamber was conditioned with the optimized a-Si:H material in order to achieve stable plasma conditions prior to the actual core layer deposition. After that, the a-Si:H core layer was deposited with an argon/silane plasma at a temperature of $300°\mathrm{C}$. The refractive index $n$ and the thickness $t$ of the a-Si:H layers were controlled with ellipsometry at 1534 nm resulting in highly uniform photonic substrates with $n=3.48$ and $t=200\mathrm{nm}$.¹⁶

The absorption loss of the optimized a-Si:H material has been already reported to be lower than $0.5\mathrm{dB}/\mathrm{cm}$ at 1550-nm wavelength.³ The photonic wires and the disk resonators were patterned with positive photoresist and electron beam lithography at Heinrich Hertz Institute, Berlin. In order to smooth the photo resist and to minimize line-edge roughness at the photonic wire sidewalls, a short descum process with ${\mathrm{O}}_2$ plasma with 80-W coil power was applied. The a-Si:H core layer was then structured by highly anisotropic dry etching process with a continuous flow of ${\mathrm{C}}_4{\mathrm{F}}_8$ and ${\mathrm{SF}}_6$ in an ICP, which allows for smooth and straight sidewalls. The residual resist was removed by an oxygen plasma, and the wafer was fragmented into photonic chips, which were cleaned with wet chemicals and were left uncladded for the optical measurements.

3.2.

Optical Measurement Setup

The microdisk and microring spectra were characterized with optical transmission measurements. Closed-loop controlled piezo-electric alignment stages and different cameras were used to maximize the coupling to the photonic chip. The photonic sensors were placed on a thermo electric controller (TEC) temperature stabilized chip mount such that a constant temperature with variations less than $0.01°\mathrm{C}$ could be maintained during the experiments. In case of the disk resonators, a lensed fiber was used to launch the NIR-light (Agilent 8164B) into the cleaved facets of the supporting photonic wire waveguides. A microscope objective with high numerical aperture of $\mathrm{NA}=0.65$ was utilized to collect the output signal and guide it to a fiber-coupled collimator connected with an InGaAs photodetector. A polarization synthesizer was used to set the qTM-polarization of the probe light and a polarizing cube was introduced into the outcoupled light beam in order to ensure the polarization settings, thereby achieving a polarization extinction ratio of $\ge 20\mathrm{dB}$ during the experiments.

In case of the ring resonators, qTE-polarization selective grating couplers were used in order to couple the light into the photonic sensor chips and polarization paddlers were applied to maximize the coupled probe signal. During the measurements, the NIR laser was continuously swept over the desired wavelength range and the transmitted intensity was datalogged.

3.3.

Photonic Sensor Characterization

The homogeneous sensing experiments were performed with NaCl solved in DI water in order to characterize the fabricated evanescent field refractive index sensors. First, a reference measurement with pure DI water was carried out and then liquids with different percentile mass concentrations of NaCl were dropped onto the resonators such that the chips were homogeneously covered with a thin film of the analyte. The same resonators were used for all measurements and were carefully cleaned with acetone, isopropanol and rinsed with DI water after each experiment. The measurement results and the sensitivities per refractive index unit are presented in Fig. 4. The graphs show that a 1% NaCl concentration cause a resonance wavelength shift of about 0.8 nm for the 10-μm disk and 84 pm for the 10-μm ring resonator, respectively. The refractive index change of 1% NaCl that is diluted with DI water is reported to be 0.0017 per mass percent at $20°\mathrm{C}$, which results in a refractive index change of 1.3105 to 1.32765 in the given range (0% to 10%).¹⁷ From the measurements, the intrinsic sensitivities were determined to be $S_{\mathrm{RIU}}=460$ and $S_{\mathrm{RIU}}=50\mathrm{nm}/\mathrm{RIU}$. The mean $Q$-factors of the resonances that were used to determine that the sensitivities were $\approx {10}^4$ which results in a resonator-based intrinsic LoD for homogeneous sensing of $\mathrm{LoD}=3.3\times {10}^{-4}$ for the disk and $\mathrm{LoD}=3.2\times {10}^{-3}$ for the ring resonator at 1550-nm wavelength, respectively. It should be noted that this intrinsic LoD is only attributed to the resonator linewidth and can be greatly improved by the readout system (typically by a factor of ${10}^2-{10}^3$). The wavelength sensitivity of our setup was determined via the linear thermo-optic effect by applying small temperature variations by the TEC-based chip mount that resulted in $\le \pm 0.1\mathrm{pm}$ resolution after Lorentz-fitting the resonance peak with a quality factor of $Q=8500$ and an 1 h warm-up phase of the entire equipment as presented in Fig. 4(d).

Fig. 4

Measurement results of the resonance peak wavelength shift according to 1.25%, 2.5%, 5%, and 10% of NaCl-dispersed in deionized (DI) water: (a) typical drop port signal of a 10-μm ring resonator in add/drop configuration.(b) Sensitivity of the microdisk resonator for a qTM whispering gallery mode. (c) Sensitivity of the ring resonator for qTE-mode. (d) Detection limit of the measurement setup determined by the linear thermo-optic effect induced resonance peak shift.

The surface sensing experiments were performed with functionalized a-Si:H ring resonators with a radius of 20 μm, a FSR of about 4.5 nm, and a $Q$-factor of 7000 in aqueous solutions; more details about similarly fabricated resonator devices on the a-Si:H material platform are given elsewhere.¹⁸ First, a reference measurement is performed with a functionalized chip and the sensor surface covered with pure DI water. Then, the BSA proteins dissolved in DI water solution with a concentration of $1\mathrm{mg}/\mathrm{ml}$ were used to homogeneously cover the whole chip. The sensor is exposed to this solution ($\approx 100\mu \text{l}$) for 1 h and subsequently cleaned with DI water to wash away any nonbound molecules. In order to keep the same process conditions as for the reference experiment, the chip is again covered with DI water and the spectral response is measured. The overall wavelength shift was determined to be 1.25 nm. The results are shown in Fig. 5. From finite element method simulations, such a wavelength shift corresponds to a mean molecular layer thickness ($t_L$) of $\approx 8\mathrm{nm}$ bound on the resonator surface. The simulations were performed with a thin molecular layer with a refractive index of $n=1.45$, and the other material properties as reported in Sec. 2. With the given molecular layer thickness $t_{\mathrm{ml}}$, the mass of this layer $M_{\mathrm{ml}}$ can be calculated according to

Fig. 5

Measurement results of a 20-μm ring resonator covered with BSA proteins. An AFM picture of the immobilized surface and a light microscope picture of the resonator are inset.

The experimental results of Fig. 4 clearly prove that for homogeneous sensing experiments, it is more reasonable to perform the measurements with a weakly confinent resonant mode as is the case for the qTM mode, which has a stronger interaction with the analyte and consequently results in a higher sensitivity; in this case, the improvement is nearly a factor of 10. For surface sensing experiments, the main interaction of the evanscent field happens in a thin molecular layer sticked close to the sensor surface so that here a weak confinement does not necessarily improve the performance. For convenience to use the grating couplers (designed for qTE-mode) instead of cleaving the chip facets, the measurements were performed with the qTE-mode. The FEM simulations indicate that the qTM-mode would yield a 1.6 times lower detection limit. From the recent results, we suggest that a-Si:H is an attractive photonic material for lab-on-chip sensing applications, especially when combined with microfluidic channels supporting a well-defined analyte flow rate and if functionalized markers with receptors are employed which facilitate a selective detection of target molecules. The experiments show that a procedure of piranha etch, aminosilanization, and glutaraldehyde functionalization is process-compatible with low-loss amorphous silicon material, and the results from the optical characterization of a ring resonator with BSA protein adsorpted to the ring surface reveal that molecular masses down to the fg region or a surface coverage of $8.5{\mathrm{pg}/\mathrm{mm}}^2$ can be detected for a conservative assumption of a 1-pm wavelength accuracy of the readout system. Further improvements are expected by a refinement of the waveguide cross-section, an improvement of the $Q$-factor by employing less strong absorbing buffer solutions and by an optimization of the resonator functionalization procedure.