An evanescent field sensor to identify materials by contact has been demonstrated using a 3D coupled waveguide array. The array is formed by imbedding layered silicon nitride stripes as waveguide cores in polymer cladding and the top cladding layer is etched open for material sensing. When objects with different refractive indexes are placed on the surface of the sensor, the evanescent field is disturbed and both the local modal distribution and the coupling condition with the connecting segments are altered, leading to different interference patterns when light from the output facet is captured and focused onto a camera. We have chosen four conventional materials for test: polymer, silicon, aluminum and silver. The sensor is able to tell them apart with distinctive patterns. In addition, the sensor can identify the location of the contact, once the material is recognized. This simple and low-cost device, supported by the recent development of image recognition technology, may open up new possibilities in chip-based sensing applications.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Photonic integrated circuits (PICs) have continued to attract attention in recent years for lab-on-a-chip applications, due to their compact size, high sensitivity, scalable performance, low power consumption and electromagnetic immunity [1–5]. PIC-based sensors have been widely used in many fields such as chemical detection , clinical diagnosis , gas measurements, environmental monitoring [8,9], etc., as a complimentary technology to fiber optic sensors [10,11]. Among these sensors, various guided-wave structures have been adopted, such as slot waveguides , micro-ring resonators , photonic crystals , Bragg gratings [15,16], etc. The sensing mechanism relies on the modulation of the light field between the waveguide and the environment. When there is a change in the refractive index near, on or in the optical structure, the guided-wave properties such as phase, amplitude, resonant frequency, propagation direction will be altered, resulting in a shift of optical response at the receiver end.
These techniques, though novel and advantageous in many aspects, require either accurate structures and challenging fabrication technology [6,12–14,17,18], or rely on complex demodulation schemes involving spectral analysis [4,7–9,19]. In this work, we propose a material contact sensor based on three-dimensional (3D) arrayed waveguides. The fabrication needs only standard deposition technology and one-time photolithography. The identification of different materials is accomplished by recognizing different interference patterns formed by the coupling-disturbed waveguide array. No spectral analysis is needed. The concept is first proven by numerical simulations and then verified by experiments. Application aspects and plans for further development are discussed in the end.
2. Sensor design
The schematic layout of the material contact sensor is shown in Fig. 1. Multiple layers of thin silicon nitride (SiNx) stripes are imbedded in a cladding material. The top surface is etched open for direct contact with probes. The SiNx stripes work as waveguide cores and are closely packed to ensure sufficient field overlap.
The 3D coupled waveguides can be treated as a multimode system. Upon light injection from an input fiber, several eigenmodes can be excited and they propagate in the array with different phase velocities, thereby generating an interference pattern, similar to a multimode interference device . The total light field E(x, y, z) can then be expressed by
When the probe touches down, e.g., onto Region 2 in Fig. 1, the local waveguiding condition is changed, i.e., the number of supported modes, their individual eigenmode distribution ϕv(y, z) and propagation constants βv are subject to change due to the presence of the material probe as a new cladding. At the interface between Region 1 and Region 2 (x = x1), the total light field E (x1, y, z) at the end of Region 1 excites the new eigenmodes supported by Region 2 with respective coupling coefficients. These modes propagate further inside Region 2 and the excitation/coupling repeats at the interface between Region 2 and Region 3 (x = x2). Finally, the waveguides are cut open at the end of Region 3 and the total light field is focused onto an infrared camera for analysis.
From Eq. (1) and (2), it can be drawn that the type of the material, as an added cladding in Region 2, will result in an altered set of modes, therefore changing the final interference image. However, the size and position of the probe play an important role as well. In this work, we prepare the samples with the same size and place them on the same location of the chip. The purpose is to demonstrate the ability to detect different materials as proof of concept. The detailed investigation on the influence of probe size and detection of contact positions will be done in our future work.
Though in principle single-layer waveguide array can also achieve the sensing function, the variation in the intensity profile stays along one horizontal line and becomes difficult to differentiate. Our initial attempts using single-layer structures have led to interference patterns with weak changes regarding different material probes and the results are not very convincing. The multilayer structure expands the intensity distribution vertically, allowing more well-separated “spots” to be identified and compared, and can thus simplify the analysis. The design is straightforward and the fabrication, though requiring 3 times deposition technology, needs only one-time patterning process.
We choose to develop the sensor chip on polymer platform, as the fabrication process is simple, fast, and flexible. Various materials and optoelectronic components can be integrated easily on this platform . Many devices have hence been developed for applications in optical communication [20,21–23]. Low-loss SiNx waveguides have also been demonstrated on this platform . The waveguide facets can be cut open with a dicing saw and well imaged onto a camera without any polishing process .
Commercial software (Lumerical) is used to simulate the waveguide structure. Horizontally, the waveguides have the same width (w) and separation (d), while vertically they are designed to have different thickness (h1, h2, h3) and gaps (t1, t2), in order to induce a more spread-out light field. In our design, the following parameters are taken after a few rounds of optimization to maximize the mode field area for the first three TM modes supported by the array structure: w = 3.3 µm, d = 3.5 µm, h1 = 100 nm, h2 = 100 nm, h3 = 200 nm, t1 = 3.5 µm, t2 = 1 µm. The refractive index of the SiNx layer is measured by an ellipsometer and found to be 1.9 at 1550 nm. The chosen polymer cladding has an index of 1.45 at this wavelength. The propagation length of the waveguides is 5.4 mm.
According to Eq. (1) and (2), the formation of the coupled waveguide interference pattern depends on the launch condition of the initial modes. We define the plane (X=0) as the left facet of the waveguide chip and the line (Y=0, Z=0) as the central axis of the 5th (middle) waveguide with the 200-nm-thick SiNx core (top), as marked in the revised Fig. 1. We have placed the input fiber at the left facet of the sensor chip, but with an offset of 3.5 µm in Y and −3 µm in Z direction relative to the coordinate origin, respectively, aiming to achieve a balanced light spot distribution for the undisturbed case. In particular, we focus on the less-confined TM polarization, as the mode profile has a broader distribution , allowing larger evanescent field in air to interact with the probes.
Figure 2(a) plots the simulated field distribution of the TM mode (Ez component) at end of the 5.4-mm propagation length (in X) on the facet of the chip (YZ plane), for the initial, undisturbed case. The SiNx cores are arranged in a 9 × 3 array. To simulate the probe contact, we have placed a block of material (polymer and aluminum) with 200-nm thickness (Z) and 200 µm (X) × 80 µm (Y) lateral size in the middle of the top surface. The polymer material used as a probe is the same as the cladding (ZPU450 from ChemOptics). The optical parameters of aluminum are taken from the database provided by the software. The field distributions are shown in Fig. 2(b) and Fig. 2(c), respectively. Clear difference can be seen from the images, especially in the segmented 9×2 blocks, proving the feasibility of the proposed sensor concept.
It should be noted that the sensor works in a different way as our previous work, where a tactile sensor is developed by imbedding a multimode fiber in a silicone pad . The fiber is flexible. The mode profiles and the interference patterns are altered by structural deformation and induced stress. In this study, however, the waveguide chip is rigid, the interference condition is changed by disturbing the waveguide surroundings, i.e., by placing another material as part of the top cladding.
Similar to the tactile sensor , we need to first test the stability of the sensor system. For repeatable contact operations of the same material, the sensor should give ideally the same interference patterns. It should also be able to return to its initial state after the probe is removed. For the stability test, we first define the averaged intensity value (AIV), as expressed in Eq. (3), to calculate the absolute change of the pixel gray levels from an image, compared to a fixed reference image. The AIV value is used as a simple measure to evaluate the stability / shift of the image patterns .
3. Device fabrication
The fabrication process is similar to that described in . The detailed steps are illustrated in Fig. 3. The polymer cladding layer is formed by spin-coating and subsequent UV curing. The fine adjustment of the cladding thickness is facilitated by low-rate reactive ion etching (RIE). The SiNx layer is deposited by plasma enhanced chemical vapor deposition at 200 °C. The certified degradation temperature of the polymer material is 300 °C. The process is repeated until all layers are formed. Since the waveguides have the same lateral width and separation, they are defined by one-time lithography, followed by Ti-deposition as hard mask, lift-off and RIE process. The etched gaps are backfilled with polymer cladding material and etched again until the top SiNx cores are exposed. No planarization or polishing step is applied. Finally, the wafer is diced into bars/chips for measurement with a standard sawing machine.
The top view of the fabricated 3D arrayed waveguides is shown in the Fig. 4(a). The cross-section of the waveguide structure is shown in Fig. 4(b), indicating the three-layer SiNx cores imbedded in polymer. The facet is not polished and the microscope is top-illuminated to highlight the thin stripes.
4. Experimental setup
Figure 5 shows the experimental setup. To test the performance of the sensor, a laser at 1550 nm is firstly regulated to TM polarization by a three-wheel polarization controller (PC) and a polarizer in front of the camera. The coupling position of the fiber and chip is adjusted until a well-distributed interference pattern is found on the camera. The fiber is then fixed onto the input facet of the chip by applying index-matching glue and subsequent UV curing. After glue fixing, the interference pattern at the output end of the waveguide does not change, even after we remove the sample with the fiber and place it back on the stage multiple times. On the output side, an imaging system is built, consisting of a series of lenses, an adjustable iris, and a rotatable polarizer. The interference pattern from the waveguides is then focused onto an infrared camera (Xenics Bobcat-640-GigE).
Four material samples are prepared for test, namely polymer (ZPU450), bare silicon chip (Si), silver (Ag) and aluminum (Al). The polymer sample is made by spin-coating and curing a layer of ∼ 8-µm thick ZPU450 on a silicon chip. Aluminum and silver come in foil form of ∼ 20 µm and 0.3 µm thickness, respectively. They are rolled flat and glued on silicon chips. All samples have the same area (1 mm × 1 mm). Aluminum and silver can also be deposited on the surface of the silicon chip by standard evaporation process.
The samples are attached to a flat-head spring probe with a diameter of 1 mm as shown in Fig. 6(a). The spring probe with the material sample is then mounted on a 3-axis motor stage with a positioning accuracy of ±2 µm. A control program is developed to steer the motor stage automatically. Laterally, the sample is placed onto the chip with the help of the on-chip markers. Vertically, the probe is first lowered down close to the surface of the chip and then touched down in step of 10 µm. The interference pattern on the camera is monitored at the same time until visible change is observed. The probe continues to touch down until there is no more changes in the interference pattern. The material sample is then believed to be in good contact with the chip. The interference pattern is then recorded for labelling and analysis.
The initial, undisturbed interference image is shown in Fig. 6(b). The difference between Fig. 6(b) and the simulation results of Fig. 2(a) can be attributed to the structural deviation from fabrication imperfections as well as the variation of coupling positions from the input fiber to the chip.
5. Results and analysis
5.1 System stability test
In our laboratory, the room temperature is controlled at 21 °C ± 1 °C, but the chip itself has neither temperature stabilizing unit or any cover to protect it from humidity and dust. We have left the laser on for 12 hours and taken the interference pattern of the chip without any probe attachment every 10 minutes. The AIVs of the images are calculated, relative to the first image, and the changes are shown in Fig. 7(a). The maximal fluctuation is less than 1%, indicating a good system stability of the experiment setup.
When a probe is applied, we record the value in the Z axis of the motor stage when sufficient contact is reached. The probe is lifted up / released and lowered / touched down again at the same Z value. The interference images for both lift-up and touch-down cases are saved. The process is repeated 40 times and the AIV of the images are calculated, relative to the first two images, respectively. The changes of the AIVs are displayed in Fig. 7(b). The fluctuation is well within 0.5%, showing that the sensor chip is durable for repeatable operations with our touch-down method.
5.2 Identification of different materials
After the samples and probes are prepared, they are mounted on the motor stage, placed at the same target position laterally, and touched down for sensing, as shown in Fig. 8(a). The interference patterns are captured and displayed in Fig. 8(b) for the four different materials: ZPU450, Si, Al and Ag.
To mark the difference more clearly, we have divided the image into 9×2 blocks, indicated by the red lines in Fig. 8(b). The patterns are distinctive, compared to the reference pattern without probe contact, as shown in Fig. 6(b). One simple way to label the different materials is to reduce the 2D image into a vector of local intensity integrals. We have summed up the camera response counts in each of the 18 blocks from one interference image, normalized to the maximal gray scale, and plotted them in Fig. 9 for the four materials. The 18 values can then be recorded for each material and used as reference for detection.
5.3 Dependence of probe position
The coupling among the waveguides and the total light field depends also on the location of the external disturbance. From the simulations, we have obtained different interference patterns when the probe is placed at different positions on top of the sensor chip. To verify this experimentally, we have prepared the probe with ZPU450 material and placed it at 4 different locations for contact, as indicated in Fig. 10(a). The corresponding images are shown in Fig. 10(b).
Using the same method, we have summed up the intensity counts in each of the blocks and plotted them in Fig. 11. The values are also distinctive. This technology can be used to develop a tactile sensor, similar to the fiber-based solution . The detection accuracy can be much higher, as the waveguides are more closely packed and can tell the touch position difference in micrometer scale. Furthermore, the interrogation method based the comparison of a vector of merely 18 values can be much simpler and faster than the image recognition program based on convolutional neural networks .
To distinguish different materials, the obtained 18-value vector from an unknown sample can be compared with the vector of known materials in the library. Their differences can be quantified by the mean square error (MSE). The smaller the MSE value, the closer is the material to that reference. The same method can be applied for the recognition of the probe position.
To conclude, we have demonstrated a material contact sensor based on 3D coupled waveguide array. Three-layer SiNx stripes are imbedded in polymer cladding to render a spread-out light field distribution for sensing and formation of interference images for detection/analysis. The sensor chip is fabricated with mature, low-temperature, and low-cost technology. When placed at the same location on the top surface, the sensor is able to tell four materials apart with distinctive patterns. To facilitate the interrogation, the image is reduced to a vector of 18 values, summing up the intensity counts in the 9×2 blocks on the camera area. When the probe material is fixed, the sensor is also able to tell the contact position and can be further developed into precise tactile sensors. We believe this technology of using optical waveguides and direct image analysis for material and tactile sensing could open up possibilities in developing miniature sensor systems for robotics and other important application areas.
As future work, we plan to expand the library of detectable materials and investigate the limitation on detection accuracy, e.g., finding the fine difference between materials with similar refractive indexes. We also aim to scale up the waveguide array for the precise, 2D detection of contact positions. The interrogation method based on data down-scaling will also be studied for comprehensive and reliable detection of materials and touch positions simultaneously.
We thank Hao Yu from China Academy of Art for the help in figure art design.
The authors declare no conflicts of interest.
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
1. M. G. Scullion, T. F. Krauss, and D. Falco, “A Slotted Photonic Crystal Sensors,” Sensors 13(3), 3675–3710 (2013). [CrossRef]
2. F. Dell’Olio and V. M. Passaro, “Optical sensing by optimized silicon slot waveguides,” Opt. Express 15(8), 4977–4993 (2007). [CrossRef]
3. B. Kumari, R. K. Varshney, and B. P. Pal, “Design of chip scale silicon rib slot waveguide for sub-ppm detection of N2O gas at mid-IR band,” Sens. Actuators B: Chem 255, 3409–3416 (2018). [CrossRef]
4. A. Raza, S. Clemmen, P. Wuytens, M. de Goede, A. S. K. Tong, N. Le Thomas, C. Liu, J. Suntivich, A. G. Skirtach, S. M. Garcia-Blanco, D. J. Blumenthal, J. S. Wilkinson, and R. Baets, “High index contrast photonic platforms for on-chip Raman spectroscopy,” Opt. Express 27(16), 23067–23079 (2019). [CrossRef]
5. H. Nazemi, A. Joseph, J. Park, and A. Emadi, “Advanced Micro- and Nano-Gas Sensor Technology: A Review,” Sensors 19(6), 1285 (2019). [CrossRef]
6. Y. Chen, Z. Li, H. Yi, Z. Zhou, and J. Yu, “Microring resonator for glucose sensing applications,” Front. Optoelectron. China 2(3), 304–307 (2009). [CrossRef]
7. Z. Wang, Y. Liu, C. Gong, Z. Yuan, L. Shen, P. Chang, K. Liu, T. Xu, J. Jiang, Y. Chen, and T. Liu, “Liquid crystal-amplified optofluidic biosensor for ultra-highly sensitive and stable protein assay,” PhotoniX 2(1), 1–16 (2021). [CrossRef]
8. C. Kraeh, J. L. Martinez-Hurtado, A. Popescu, H. Hedler, and J. J. Finley, “Slow light enhanced gas sensing in photonic crystals,” Opt. Mater 76, 106–110 (2018). [CrossRef]
9. Y. Tomono, H. Hoshi, and H. Shimizu, “CO2 Detection with Si Slot Waveguide Ring Resonators Toward On-chip Specific Gas Sensing,” in 2019 Conference on Lasers and Electro-Optics (CLEO), 1–2 (2019).
10. X. Bao, Z. Zhou, and Y. Wang, “Review: distributed time-domain sensors based on Brillouin scattering and FWM enhanced SBS for temperature, strain and acoustic wave detection,” PhotoniX 2(1), 1–29 (2021). [CrossRef]
11. C. Zhu and J. Huang, “Sensitivity-enhanced microwave-photonic optical fiber interferometry based on the Vernier effect,” Opt. Express 29(11), 16820–16832 (2021). [CrossRef]
12. J. Witzens, T. Baehr-Jones, and M. Hochberg, “Design of transmission line driven slot waveguide Mach-Zehnder interferometers and application to analog optical links,” Opt. Express 18(16), 16902–16928 (2010). [CrossRef]
13. R. Jannesari, C. Ranacher, C. Consani, T. Grille, and B. Jakoby, “Sensitivity optimization of a photonic crystal ring resonator for gas sensing applications,” Sensors Actuators A Phys 264, 347–351 (2017). [CrossRef]
14. C. Fenzl, T. Hirsch, and O. S. Wolfbeis, “Photonic crystals for chemical sensing and biosensing,” Angew. Chem. Int. Ed. 53(13), 3318–3335 (2014). [CrossRef]
15. Y. Singh, A. Sadhu, and S. K. Raghuwanshi, “Development and experimental analysis of titanium dioxide (tio2) coated etched fiber bragg grating sensor for chemical sensing,” IEEE Sens. J. 20(15), 8528–8534 (2020). [CrossRef]
16. C. Lin, C. Liao, J. Wang, J. He, Y. Wang, Z. Li, T. Yang, F. Zhu, K. Yang, Z. Zhang, and Y. Wang, “Fiber surface Bragg grating waveguide for refractive index measurements,” Opt. Lett. 42(9), 1684–1687 (2017). [CrossRef]
17. R. R. Singh, N. Malviya, and V. Priye, “Parametric analysis of silicon nanowire optical rectangular waveguide sensor,” IEEE Photonics Technol. Lett. 28(24), 2889–2892 (2016). [CrossRef]
18. M. Vlk, A. Datta, S. Alberti, H. D. Yallew, V. Mittal, G. S. Murugan, and J. Jágerská, “Extraordinary evanescent field confinement waveguide sensor for mid-infrared trace gas spectroscopy,” Light: Sci. Appl. 10(1), 1–7 (2021). [CrossRef]
19. E. Ryckeboer, R. Bockstaele, M. Vanslembrouck, and R. Baets, “Glucose sensing by waveguide-based absorption spectroscopy on a silicon chip,” Biomed. Opt. Express 5(5), 1636–1648 (2014). [CrossRef]
20. Z. Dang, T. Chen, Z. Ding, Z. Liu, X. Zhang, X. Jiang, and Z. Zhang, “Multiport all-logic optical switch based on thermally altered light paths in a multimode waveguide,” Opt. Lett. 46(13), 3025–3028 (2021). [CrossRef]
21. Z. Zhang, N. Mettbach, C. Zawadzki, J. Wang, D. Schmidt, W. Brinker, N. Grote, M. Schell, and N. Keil, “Polymer-based photonic toolbox: passive components, hybrid integration and polarisation control,” IET Optoelectron. 5(5), 226–232 (2011). [CrossRef]
22. D. de Felipe, Z. Zhang, W. Brinker, M. Kleinert, A. M. Novo, C. Zawadzki, M. Moehrle, and N. Keil, “Polymer-based external cavity lasers: tuning efficiency, reliability, and polarization diversity,” IEEE Photonics Technol. Lett. 26(14), 1391–1394 (2014). [CrossRef]
23. V. Katopodis, C. Kouloumentas, A. Konczykowska, F. Jorge, P. Groumas, Z. Zhang, A. Beretta, A. Dede, J.-Y. Dupuy, V. Nodjiadjim, G. Cangini, G. V. Büren, E. Miller, R. Dinu, J. H. Choi, D. Pech, N. Keil, H.-G. Bach, N. Grote, A. Vannucci, and H. Avramopoulos, “Serial 100 Gb/s connectivity based on polymer photonics and InP-DHBT electronics,” Opt. Express 20(27), 28538–28543 (2012). [CrossRef]
24. Z. Zhang, D. Liu, D. de Felipe, A. Liu, N. Keil, and N. Grote, “Polymer embedded silicon nitride thermally tunable Bragg grating filters,” Appl. Phys. Lett. 102(18), 181105 (2013). [CrossRef]
25. Z. Ding and Z. Zhang, “2D tactile sensor based on multimode interference and deep learning,” Opt. Laser Technol. 136, 106760 (2021). [CrossRef]
26. P. Wu, S. Zhu, M. Hong, F. Chen, and H. Liu, “Specklegram temperature sensor based on femtosecond laser inscribed depressed cladding waveguides in Nd: YAG crystal,” Opt. Laser Technol 113, 11–14 (2019). [CrossRef]