Abstract
Introduction
With the development of nanofabrication technologies, an increasing effort has been placed on plasmonic sensors [1,2]. Plasmonic sensors are based on surface plasmon resonance (SPR) or localized surface plasmon resonance (LSPR) [1-3]. When the target analytes bond to the functionalized material and altered the local refractive index, the resonance wavelength shifts due to the change of the local refractive index. Many types of plasmonic sensors have been studied and developed for chemical sensing or biochemical sensing applications in the last decade. There are two main challenges for chemical and biochemical sensing, i.e., sensitivity and specificity, especially when working in sensing applications in a complex background such as environmental monitoring, breath analysis, hazard tracing, etc.[1,3,4]. Besides, the development of artificial intelligence and machine learning technology provides the feasibility to correlate sensor responses to multiple parameters[4]. In this work, we report a novel plasmonic sensing platform with enhanced sensitivity and an AI-assisted signal processing algorithm to improve the specificity of plasmonic sensors.
Sensor Simulation
To achieve higher sensitivity and a broader range of detection. We have theoretically studied and simulated a sensing platform, namely, nano plasmonic pillars (NPP) (Figure 1). SiO2 nanopillars with a diameter of 100 nm are coated with a 20 nm-thick gold layer. The nanopillars are hexagonally displaced with a displacement of 300 nm. The Au-coated nanopillar has an interface between metal (Au) and insulators (SiO2) along the nanopillars' whole surface area. This structure will generate localized surface plasmon resonance (LSPR) along the nanopillars and improve the sensitivity due to the large effective surface area. The parameters, i.e., size, material, and displacement of the nanopillars, are simulated and optimized using finite element analysis. Figure 2 compares the simulated wavelength shift of the Au-coated SiO2 pillars and the ones of solid Au pillars when the local refractive index changes from 1.0 to 1.05. It is shown that the sensitivity (slope of the fitted function) of the simulated design is 2.90 times higher than the solid pillar.
AI-Assisted Signal Processing
One advantage of plasmonic sensors is that the sensor response can be captured by CMOS cameras. In this way, the wavelength shift caused by the analytes can be monitored as a color change of the captured image. With the rapid development of machine learning technology, convolutional neural networks have become a robust tool to help process the sensor signals. In this work, we developed a convolutional neural network-based algorithm that can successfully predict the local refractive index based on the simulated wavelength. Figure 3 shows the signal processing method applied in this work. First, the transmission spectra of the NPP with refractive indices of 1.0, 1.01, 1.02, 1.03, 1.04, and 1.05 are simulated. The spectra have dual peaks located in the visible wavelength due to the specifically designed plasmonic features (Figure 1). Second, approximate RGB values are calculated based on the spectra following Plank's law. A random integer (from -3 to 3) was added to the calculated RGB as random noise. Over 5,000 images (64x64 pixels each) for each corresponding spectrum (30,000 images in total) were generated based on the RGB value and the random noises. Finally, the images were trained and tested by the developed convolutional neural network. We successfully predicted the local refractive index with 99% accuracy with 5 cross-validations.
Note: we use the simulated images to prove the concept in this work. The same method can be applied when detecting real analytes with proper adjustment. We believe the reported work can help to improve both sensitivity and selectivity in chemical and biochemical sensing.
References
[1] Soler, M., Huertas, C. S., & Lechuga, L. M. (2019). Label-free plasmonic biosensors for point-of-care diagnostics: a review. Expert review of molecular diagnostics, 19(1), 71-81.
[2] Xu, Y., Bai, P., Zhou, X., Akimov, Y., Png, C. E., Ang, L. K., ... & Wu, L. (2019). Optical refractive index sensors with plasmonic and photonic structures: promising and inconvenient truth. Advanced Optical Materials, 7(9), 1801433..
[3] Zhao, Y., Mukherjee, K., Benkstein, K. D., Sun, L., Steffens, K. L., Montgomery, C. B., ... & Zaghloul, M. E. (2019). Miniaturized nanohole array based plasmonic sensor for the detection of acetone and ethanol with insights into the kinetics of adsorptive plasmonic sensing. Nanoscale, 11(24), 11922-11932.
[4] Feng, S., Farha, F., Li, Q., Wan, Y., Xu, Y., Zhang, T., & Ning, H. (2019). Review on smart gas sensing technology. Sensors, 19(17), 3760.
Figure 1