Plasmonic materials and their applications in developing surface-enhanced Raman biosensors

The work described herein outlines the steps taken toward engineering plasmonic structures for spectroscopic detection and quantification of bioanalytes. The sections will focus on three aspects of the development process: fabrication and optimization of plasmonic materials, application of these in developing a biosensing platform based on surface-enhanced Raman spectroscopy (SERS), and development of robust mathematical models for sensor calibration.;When light interacts with nanoparticles, they exhibit localized surface plasmon resonance (LSPR) which depends on their shape, size, and dielectric environment. The LSPR produces a localized electromagnetic field at the surface, which enhances the Raman scattering at the surface by as much as 10 6 -- 108. By varying the size of the spheres mask used to fabricate the arrays of nanoparticles, we can tune the LSPR frequency to match the Raman excitation wavelength to achieve maximum enhancement. We explore two types of SERS-active plasmonic nanostructures, Ag film over nanospheres (AgFON) and Ag nanoparticles in nanowells.;The SERS sensor utilizes AgFON surfaces for detection of glucose. AgFONs were functionalized with self-assembled monolayers (SAMs) to increase the affinity of glucose to the surface. The sensor performance was evaluated based on the following characteristics: reversibility, time response, stability, and accuracy of measurements. Furthermore, we demonstrated that the same sensor can be used for lactate detection, exhibiting the similar characteristics as the glucose sensor. Simultaneous detection of glucose and lactate was demonstrated. In an effort to demonstrate the feasibility of the SERS sensor to as a multianalyte sensing platform, we have also shown that we can directly measure tripalmitin at physiological concentrations.;Finally, we develop a new robust approach for multivariate calibration where the system is subject to perturbations. Robust methods are used for worst case scenario predictions by defining an uncertainty set for the data. We adapt this method to generate a minimum and a maximum prediction for a given sample, producing a range prediction rather than a single value prediction. This calibration model based on robust least squares was applied to a set of SER spectra generated by the glucose sensor.