All-optical on-chip sensor for high refractive index sensing in photonic crystals

In this paper we demonstrate an optical sensor designed to detect material infiltrations with relatively high indices, based on a two-dimensional photonic crystal cavity structure. The locations and sizes of the holes surrounding a L3 cavity were modified to increase the Q factor to a value of 1500 with a high refractive index infilling of . With precise design and simulation, we overcome the difficulty of low index contrast, and observe a very clear wavelength shift of 10.4 nm in simulation and 12.4 nm in experiment between water and oil samples at resonance.

Introduction. -The developments of all-optical onchip sensors have attracted many attention in both academia and industry, especially in the field of biological and chemical detection. The most well-exploited techniques for optical sensing are based on the principle of surface plasmon [1,2], interferometry [3][4][5][6] and optical resonance [7]. The principles of operating these sensors rely on variations of their optical properties as the tested analytes change [8,9]. Sensors made of photonic crystal cavities work under the principle of optical resonance. Dips or peaks can be found in their transmission spectra at resonance. As the refractive indices are modified, wavelength shifts of the dips or peaks are traceable. With precise calibration, the shifts may even work as fingerprint identification. Photonic crystal sensors have many advantages compared to other kinds of sensors and have become one of the hot research topics. Firstly, photonic crystal has very small footprint and high quality factor/volume. It has been proven that with fine structural optimization a photonic crystal cavity can have a quality factor (Q) as high as 10 6 [10][11][12][13] in very small dimensions [14,15], on the order of λ 3 , which enhances the light-matter interaction effect but decreases the volume of interaction. Therefore, tiny variations in refractive index are measureable and reflected in the optical properties of the photonic (a) E-mail: H.Salemink@science.ru.nl crystal cavities. Secondly, photonic crystal can be easily integrated with other on-chip elements. A potential utilization of this feature is being used as a part of a lab on chip [16][17][18][19][20][21][22].
However, the widths of the transmission band gaps and the light confinement in the photonic crystal cavity strongly depend on the refractive index contrast between the solid material and periodically embedded materials. For those previous works mentioned above, high Q factors were obtained under high refractive index contrast (silicon vs. air). In many practical applications, sensors are expected to work under liquid environment. A high refractive index of the liquid will unavoidably cause adverse results such as narrow band gaps, low Q factors, weak light confinement in both vertical and horizontal directions, and broad resonant peaks in transmission. These problems can seriously affect the functionalities of the sensors, lower the sensitivities or resolutions in detections.
Here we present our investigation towards the development of a sensor based on a modified L3 cavity that can overcome the above difficulties. As will be demonstrated, our sensor is designed to work under high refractive index material infilling of n = 1.5, which is much higher than many typical liquids. Theoretical analysis shows that after modifying the structure of the sensor, the Q factor and intensity are highly increased. In experiment, an oil sample with a refractive index of 1.45 is used to demonstrate the sensing ability of our sensor. We observe a wavelength shift of 12.4 nm with a refractive index variation of 0.12.
Design. - Figure 1 shows a photonic crystal cavity in a silicon-on-insulator (SOI) wafer. The investigated structure consists of holes being etched only through the top silicon membrane. Photonic crystal cavity is created by solid-filling three holes in the Γ-K direction which is usually called the L3 cavity. The confinement of light in the vertical direction is ensured by total internal refection (TIR) at the interfaces between the silicon membrane and lower refractive index materials both beneath and above the membrane. Instead of a traditional airbridge structure, silicon dioxide layer with a thickness of 2 micron is used as a bottom to support the top silicon membrane. The liquid analyte covers the top of the photonic crystal and fills in the holes. The structure shown here has a lattice constant a. Its radii of the holes r are set as r/a = 0.39. The thickness h of the silicon membrane equals to h/a = 0.52 and its refractive index n slab is set as 3.4. In order to make the sensor work in the near-infrared range, we choose the parameters as follows: a = 500 nm, h = 260 nm, r = 195 nm.
The Q factor is tuned by varying the geometric parameters of the holes surrounding the cavity. According to those previous investigations on high Q cavities design, light confinement can be optimized by adjusting the profile of electromagnetic-field distribution of resonant optical modes. Enhancing the confinement in perpendicular direction [23], smoothing the field distribution at the boundaries [11], and creating band edge mismatching [13] are all efficient methods in achieving high Q cavities. Here we use those methods to increase the Q factor of our sensor. A three-dimensional finite different time domain (FDTD) simulation program, MEEP [24], is resorted to in order to examine the effect of shifting holes. All dimensions are normalized to the lattice constant a. In the first step of our design, the silicon slab is fully exposed to a fill-in material with a refractive index of n = 1.5 in our simulation. The cavity is surrounded by 10 rows of holes in the Γ-K direction, and 12 rows of holes in the Γ-M direction. The entire simulation region is surrounded by a perfectly matched layer (PML) with a thickness of 2.0a, absorbing any residual light at the boundaries [23]. The resonant frequencies and quality factors are computed by the software of harmonic inversion (the built-in "harminv" in MEEP) [25], which decouples the cavity fields into individual sinusoids and calculates their decay rates. A broadband pulse is placed inside the L3 cavity as a TE-polarized dipole point source with Gaussian frequency distribution to excite the cavity modes. The default power intensity of the source is 1, and the light source has location offsets of 0.1a in both Γ-K and Γ-M directions in order to excite all possible optical modes regardless of modes' symmetry. Resonant modes maintain high Q factors after 400 time periods after the excitation. First, we calculate the resonant modes of a typical L3 cavity without any structural modifications. The unmodified cavity has four resonant modes within the band gap, and their quality factors and intensities are shown in fig. 2(c) and fig. 2(e). Only the mode around λ = 1700 nm has an acceptable quality factor of 150 and an intensity of 0.45. We simulate the distribution of the electric field of this mode and show its E y component in fig. 2(a). Obviously, the electric field is confined not only in the cavity but also extend to the neighbouring area, including the first and the second neighbouring rows in both Γ-K and Γ-M directions. According to the theory mentioned in ref. [13], abrupt structural changes at the boundaries must be avoided to minimize leakages above the light cone. Therefore, a group of displacements were done to smooth the boundaries and suppress the leakages. We gradually relocate and change the sizes of the neighbouring holes to control the light leakages. The optimized sensor has a structure described here. The width of the cavity is modified by locally slightly shifting two neighbouring rows of holes away from the center in the Γ-M direction, with shifting distances of 0.02a and 0.04a, which are 10 nm, and 20 nm for our sensor. We also modify the radii of the shifted holes. The first neighbouring row has radius of r a = 0.72r and the second neighbouring row has enlarged radius of r b = 1.2r. Modifications are also done on the holes in the Γ-K direction. The nearest pair of holes is shifted 0.21a outwardly and reduced to r c = 0.56r in radius. The next pair also has reduced radius of r d = 0.87r.
Several resonant frequencies of such a cavity have been found within the transmission band gap. Their Q factors and intensities are shown in fig. 2(d) and fig. 2(f). For sensor applications, not only the quality factors are crucial to gain high sensitivities, but also the power intensities of trapped photons at resonance are of great importance. Therefore, we focus our attention on the resonance modes which have both higher Q factors and higher intensities. In our case, the resonant mode around 1560 nm has a Q factor higher than 1500 and its normalized accumulated intensity is 1.9. The electric-field distribution is also simulated and shown in fig. 2(b) to confirm the good light confinement.
To illustrate the operating principal of the sensor and to quantitatively estimate its sensitivity, a series of threedimensional FDTD simulations have been performed. The simulation domain comprises the cavity itself with a lattice constant of 500 nm and a pair of single-mode photonic crystal waveguides as the light input/output ports. We do the simulation in two steps. In the first step, the holes of the photonic crystal are filled with air, water and oil of which refractive indices are set to be 1.0, 1.33 and 1.45, respectively. In the second step, we make the analytes only touch the top of the sensor rather than fill into the holes in order to simulate the situation of a highly hydrophobic surface.
The calculated output spectrum of a photonic crystal sensor is demonstrated in fig. 3. Under infiltrations, the two peaks in black and red in fig. 3(a) correspond to the calculated resonances. Shift in resonant frequencies indicates a change in the refractive index of the analytes. As can be seen, there is no resonant peak for air infiltration, but the resonant peak of water infiltration starts to show, and it is quite obvious for oil infiltration. A change in the refractive index of Δn = 0.12 between water and oil results in a spectral shift of approximately 10.4 nm. It should be noticed that our sensor is designed for high index infiltration, therefore the higher index an analyte has, the higher output intensity it shows. Compared to the analytes infiltrations, output spectra of the non-infiltrations are quite different. No clear resonant peaks with good Q factors and intensities can be found in the same wavelength range in fig. 3(b). A sensor with hydrophobic surface works similarly to an air infiltration.
Fabrication. -The real photonic crystal sensor shown in this paper was fabricated on silicon-on-insulator (SOI) wafer with a 260 nm thick silicon membrane and a 2 μm thick silicon dioxide insulator layer. The fabrication of the device started from cleaning a 2 cm 2 SOI chip with nitric acid to remove any residual organics. ZEP520, a positive electron-beam resist, was spun on top of the SOI chip with an approximate thickness of 120 nm followed by a pre-baking at 175 • C for 15 min to remove any residual solvent. The photonic crystal was patterned using a Leica EBPG 5000+ e-beam lithography system operating at 100 keV. The resist was developed with a microchem 1:3 MIBK:IPA solution for 30 s followed by a 10 s IPA rinse. Exposed areas of the SOI wafer were then etched using inductively coupled plasma (ICP) dry etching system Alcatel AMS100 with gas flows of 75 sccm SF 6 and 25 sccm O 2 at −120 • C. Following the silicon etch, the remaining ZEP520 was dissolved in a PRS3000 solution, and IPA was used again to remove any surface contaminants and ensure a non-hydrophobic surface. Finally, optical waveguides were exposed for measurement by hand-cleaving at the facets.
For our photonic crystal sensor, the resonant wavelength is near λ = 1550 nm. The lattice constant was therefore set to be a = 500 nm with hole radius r = 195 nm. The sensor consists of a central cavity, 9 rows of holes on both sides along the Γ-M direction, and a pair of input and output photonic crystal waveguides. The photonic crystal waveguides are created by solid-filling a single line in the Γ-K direction. Coupling into and out of the photonic crystal waveguides are accomplished by traditional tapered strip waveguides ( fig. 4). The strip waveguides are 220 nm thick and taper from 2.5 μm wide at the chip edge to 866 nm wide at the photonic crystal waveguide interface, in order to minimize the mode mismatch.

Experiment and analysis. -A tunable laser of
Santec TSL-510 with a broadband power output from 3 mW to 10 mW and wavelengths ranging from 1440 nm to 1630 nm was used as our light source. A polarization maintained lensed fiber was used to couple light from the laser to the input strip waveguide. TE polarization was selected before the fiber was mounted onto a three-axis optical stage at an extinction ratio of 100:1. The sample holder also has three-dimensional flexibilities in fine position adjustment. A group of objective lenses focused at the facet of the output strip waveguide, which collected the infrared light emitted from the waveguide and connected to a sensitive photon detector. Analytes were placed directly on top of the sensor by syringes as shown in fig. 5. After each analyte was measured, the liquid drop was blown away from the chip simply by pressured nitrogen gas.  The measured transmission spectra of our sensor are plotted in fig. 6. As can be seen sharp resonant peaks appear in the output spectra. The appearances of these peaks are clear evidences of the actual analytes infiltrations. Shifts of the resonant wavelengths depend on the refractive indices variations. The solid spots in fig. 6 are original data gotten from experiment, and the lines are the Lorentz fittings of those spots. The black spectrum, which corresponding to water infiltration, shows a Q factor of 1518 around 1564.2 nm. The red spectrum is obtained after the oil was placed on the sensor, and its Q factor is 950 at 1576.6 nm. As expected there is no obvious resonant peak for air infiltration. We observe very good matches between experimental data and theoretical simulations. In the presence of increased refractive indices, the resonant wavelength shift is 12.4 nm for water/oil in experiment, while the simulated wavelength shift is 10.4 nm for the same index change. We calculate the sensitivity of our sensor which is defined as the wavelength shift per refractive index unit (RIU). A slight index difference of 0.12 results in a wavelength shift of 12.4 nm. The device shows a sensitivity of 103 nm RIU −1 . With all the modifications mentioned above, high Q factors can still be obtained with high refractive index infiltration to create clear wavelength shifts in the spectra. The output intensity might be further increased by enhancing the coupling between fibers and waveguides using the skill mentioned in ref. [26], which is in our future plan.
Conclusion. -We have demonstrated that a modified two-dimensional photonic crystal cavity can be used as a sensitive optical sensor for high refractive index material infiltrations. We overcome the difficulties of low index contrasts between analytes and photonic crystal, and achieve an optimized L3 cavity with both high Q factors and intensities. By sensing the refractive index changes of the analytes, clear shifts of resonant wavelengths can be observed. A slight refractive index difference of 0.12 results in a wavelength shift of 12.4 nm, which reflects a sensitivity of 103 nm per unit change in the refractive index. The sensor we designed is an all-optical sensor. This feature makes it less vulnerable to external electro-magnetic field than many other sensors such as metal-, semiconductor-, and plasmonic-based nano-sensors. We believe the development of optical nano-sensors is of great importance to environmental monitoring and biological detection. Our future plan is to further increase the sensitivity of such a high index infiltrated sensor.