Raman spectra provide rich vibrational signals that represent the fingerprint of molecules, and more importantly, such signals are insensitive to water so that recently the technique of Raman scattering emerges a promising method for environmental and biological trace analysis. Nevertheless, the scattering cross section of Raman signals is usually small, and therefore it is crucial to further enhance the signals of Raman scattering. In this work, we demonstrate a low-cost and highly sensitive surface enhanced Raman scattering (SERS) substrate, which is comprised of the silicon nanowire (SiNW) array decorated with Au nanoparticles (AuNPs) on the surface and incorporated with a layer of Au backplane at the bottom, as shown in Fig. 1. Firstly, the SiNW array was prepared by a metal-assisted chemical etching (MaCE) method as a template of the SERS substrate. Then, the gold nanoparticles (AuNPs), which were successfully decorated on the surface of SiNWs array by an electron-beam evaporator under oblique angle deposition (OAD), enable localized surface plasmon resonance (LSPR) to substantially intensify the signal of Raman scattering from the analyte. Besides, we deposited an additional Au layer at the bottom of the SiNW array by electron-beam deposition under normal incidence. This additional Au layer, termed as a metal backplane (MBP), facilitates to reflect the back-scattered field rather than being absorbed by the silicon substrate, leading to the further enhancement of the SERS signals. Finally, the performance of this SERS substrate was optimized by a statistical Taguchi method. Our experimental verification indicates that for the analyte of 10-2 M self-assembled monolayer (S.A.M.) of thiophenol molecules, our tailored SERS substrate presents the average Raman signal up to1740 counts per second under a near infrared laser excitation (785 nm), which is 1.78 times stronger than a commercialized SERS substrate (KlariteĀ®). Furthermore, our low-cost and high-sensitivity SERS substrate preforms reliably, showing a small coefficient of variation (C.V.) about 4.2%.
tjyen@mx.nthu.edu.tw