UC Irvine

Surface enhanced Raman scattering (SERS) spectroscopy is a powerful tool for identifying and quantifying complex mixtures of small molecules, such as the metabolites that indicate the state of a bacterial system. SERS frequently takes advantage of metal nanoparticles to confine light onto their surface, increasing the local electric field. This field enhancement massively increases Raman scattering, and can enable single molecule detection. In this thesis, I introduce a new nanomanufacturing technique called 2-dimensional physically activated chemical self-assembly (2PAC). The focus of 2PAC is to manufacture extremely powerful and uniform SERS sensors, with a SERS enhancement factor of 109 that has a relative standard deviation of 10% over a 1 mm x 1 mm area. This technique is characterized by electron microscopy and Raman spectroscopy to elucidate its physical origin. I show that SERS sensor’s field enhancement can be further increased by 3-fold by taking advantage of Rayleigh’s anomaly using electron beam manufactured optical gratings.

While SERS enables fantastic chemical sensing, it also increases the complexity of analyzing spectra. This is because the ligands involved in 2PAC produce their own spectral signature, and the nanostructures produce a field enhancement that decays rapidly from the hotspot in withhin gaps between nanoparticles. In order to address this complexity, I have developed machine learning techniques that greatly improve analyte concentration regressions. Using a convolutional neural network, I demonstrate quantitative sensing down to 10 fM, well into the single molecule detection regime. I use these methods to address another complex problem, identifying the state of a bacterial system through its metabolome. First, I show that by tracking pyocyanin, a metabolite of Psuedomonas aeruginosa (PA), PA can be monitored as it forms a biofilm. PA is detected in just 6 hours, well before it becomes resistant to antibiotics. Then, I demonstrate a semi-supervised method of antimicrobial susceptibility testing (AST) using SERS. This method identifies which antibiotics a bacterium is susceptible to with minimal 24-hour cell culture. AST of PA is performed in just 30 minutes with over 99% accuracy. In sum, this thesis points a way forward to better healthcare through nanotechnology and machine learning.