UC Irvine

Self-assembly utilizes thermodynamic driving forces in order to achieve molecular scale control with less waste and lower costs than lithographic methods. Thus the ability to understand reactivity between functional groups on molecular structures in confined geometries provides knowledge on how to organize building blocks such as molecules and nanoparticles (NPs) into supramolecular structures and metasurfaces using scalable methods. The overall objective of my work in understanding driving forces for forming plasmonic nanostructures on surfaces from colloidal solutions was to design plasmonic nanostructures with molecular scale control for sensor applications. I show that the resulting plasmonic nanostructures can produce unique light matter interactions, controlling localized interactions with plasmons and phonons for molecular sensors with detection limits at single molecule and single cell levels. Colloidal NPs are typically stabilized via electrostatic repulsion, and thus, chemical crosslinking in solution can lead to uncontrolled aggregation. Here, I developed a robust fabrication method using long range electrokinetic forces, oscillation field electrohydrodynamic (AC-EHD) flow, to drive chemical reactions between suspended NPs on working electrode surfaces. This process forms "nanogaps" between the NPs and enables manufacturing of surface enhanced Raman scattering (SERS) sensors with record performance. By varying field frequency, isolated NPs can be deposited onto the surface at 100 Hz. At higher frequencies, e.g., 500-1500 Hz, AC-EHD flow plays a more important role to form 2D Au nanostructures with controlled nanogap spacing. In order to fabricate plasmonic nanostructures with high density over large areas, a two step deposition was utilized to first deposit isolated NPs as seeds then apply higher frequencies to grow to clusters. The resulting SERS sensor is highly sensitive to bacterial metabolites, many which have aromatic rings. Thus the bacterial metabolic responses to environmental stimuli is robustly differentiable when using machine learning (ML) algorithms to analyze the spectral data. Studies on nutrient deprivation are performed to understand general stress response, which can be useful for process monitoring in biotechnology industries. Changing nutrient source of E. coli cultures from glucose to the less preferred xylose and sucrose, provides insight on metabolic network response. The metabolite profile of bacteria in response to antibiotic treatment differentiates resistance and susceptibility as soon as 5 min to produce a new platform for rapid antimicrobial susceptibility testing. Similarly, bacterial stress responses to toxic heavy metals were used to monitor water quality. We show that detection of arsenic ions (As3+) and chromium ions (Cr6+) is possible at the single cell level. ML analysis of the vibrational spectra of metabolites released in response to As3+ and Cr6+ exposure detects concentrations 108 times lower than those leading to cell death. Transfer learning of trained algorithms to test contaminants in tap water and wastewater were able to achieve 92% accuracy.