Enhancement of the Au/ZnO-NA Plasmonic SERS Signal Using Principal Component Analysis as a Machine Learning Approach

Akhilesh Kumar Gupta,Chao-Sung Lai,Chih-Hsien Hsu

doi:10.1109/jphot.2020.3015740

Abstract

In this work, we modeled a novel approach to enhance surface-enhanced Raman scattering (SERS) signals using principal component analysis (PCA) as a machine learning approach. Zinc oxide nanoarrays (ZnO-NAs) were synthesized using a hydrothermal method followed by zinc oxide nucleation on ITO glass substrates via an oxidation furnace at 500 °C. The surface morphology was improved by short rapid thermal annealing (S-RTA) after deposition of a gold layer via a thermal evaporator to avoid chemical contamination of the sensing surface, which is a suitable plasmonic platform for the generation of “hot spots” for SERS enhancement with fewer defects. The proposed Au/ZnO-NA SERS sensor exhibited an enhancement factor (EF) of 1.15 × 10 7 via the R6G Raman probe and excellent uniformity over the entire surface. The PCA algorithm was used to extract useful features and information from the SERS signal. The algorithm was implemented with MATLAB software (R2019a) by the multivariable analytical tool to find an enhanced signal (~3 times higher) with high uniformity, which has great potential and is applicable to a wide range of probe molecules suitable in medical, safety, and environmental applications.

Highlights

IntroductionSignal amplification and the signal-to-noise ratio play a critical role
In sensing applications, signal amplification and the signal-to-noise ratio play a critical role
We established a highly sensitive and practical method with an AuNP/Zinc oxide nanoarrays (ZnO-NAs) surface-enhanced Raman scattering (SERS) sensor platform merged with multivariate statistics as a machine learning tool for the detection of R6G molecules with an enhanced (3 times higher) signal

Summary

Introduction

Signal amplification and the signal-to-noise ratio play a critical role. PCA as a machine learning approach was used to compute multivariate statistics to identify specific diseases with a strong enhanced signal. Researchers have demonstrated a SERS platform to help ensure normal delivery without infection and premature delivery of maternal diseases Their SERS platform could detect the presence and identification of prenatal diseases with the assistance of a machine learning approach using the PCA-SVM for amniotic fluids with high sensitivity [11]. We modeled the PCA algorithm as a machine learning approach for an enhanced signal with high intensity These methods are expected to expand to different research areas in the future, as they apply to other types of spectroscopies and the spectroscopic field of deep learning and artificial intelligence data processing, and can be used for the extraction of sufficient information to identify specific diseases

Experimental Details

Plasmonic Substrate

Results

PCA Techniques for the Analysis of Raman Spectra

Conclusion