This dissertation is divided into two separate parts covering my research in two different fields of optics. Part I consists of chapters 1-3 and covers experiments on periodically poled lithium niobate while Part II consists of chapters 4-6 and covers various spectroscopic methods designed for the application of in vivo blood analysis. Chapter 1 serves as a brief introduction to periodically poled lithium niobate and its fabrication process. In chapter 2, the key results of Part I, derived from a series of experiments on poling of thinned lithium niobate, are presented. Building upon these experiments, chapter 3 concludes Part I with a study on poling of crystal ion sliced lithium niobate. Part II begins with chapter 4, which describes a spectroscopic approach for non-invasive blood analysis in vivo. In chapter 5, experiments analyzing aqueous glucose solutions with mid-infrared and Raman spectroscopy are discussed. Chapter 6 concludes this dissertation with the design and demonstration of a innovative stimulated Raman spectroscopy system.
In Part I, ferroelectric poling fabrication procedures were developed, optimized, and implemented for periodic poling of thinned lithium niobate. The free-standing samples of thickness from 500 μm down to 25 μm were thinned by chemical mechanical planarization and annealed before poling. Domain structure was investigated as a function of sample thickness using Raman, scanning electron, atomic force, and optical microscopy, and broadening of poled domains was consistently found to vary with sample thickness in a strong linear correlation. Domain broadening was reduced by 38% as the thickness of the poled sample was reduced from 500 to 25 μm. Micro Raman probe measurements showed a thickness-dependent contrast in Raman active mode intensity between poled and unpoled regions, with the thinner samples having a higher intensity contrast.
To explore poling on even thinner free-standing samples, crystal ion sliced lithium niobate thin films of 10 μm in thickness were fabricated. Chemical mechanical planarization of the ion-implanted layer and annealing was performed to prepare the thin films for poling. Ferroelectric poling of the crystal ion sliced samples was attempted, but unsuccessful, suggesting that alternative fabrication processes may be necessary for poling of crystal ion sliced thin films.
In Part II, several disparate experiments were conducted to progress towards a common overarching goal of developing a spectroscopic method for non-invasive whole blood analysis and metabolite monitoring. A portable visible and near-infrared spectroscopy system for in vivo blood spectral identification was developed and demonstrated in a clinical setting. A custom-designed clip attached the illumination and collection optics to opposite sides of the patients’ fifth fingertip, and applied gentle pressure, gradually pushing a small quantity of blood away from the measurement site, and inducing a time-dependent change in the effective path length of blood. Time-dependent visible and near-infrared spectra were measured from the collected transmitted and scattered light. A maximum likelihood model was developed to leverage the time-dependent spectral component and identify the spectrum of blood, isolating it from that of surrounding tissue.
A second set of experiments were conducted to develop a model for predicting glucose concentrations from measured mid-infrared transmission and spontaneous Raman scattering spectra. Partial least squares regression models were trained, validated, and tested on the spectra of aqueous 0-10 mM glucose solutions measured by both spectroscopic modalies. The models proved to be accurate predictors of glucose concentration as the mean squared error of the model based on mid-infrared spectra ranged from 0.10 - 0.74 mM, and that of the Raman-based model ranged from 0.26 - 0.93 mM.
Finally, an LED-based stimulated Raman system was innovated to improve upon the relatively weak spontaneous Raman signal in a cost-effective manner. Stimulated Raman gain using a broadband LED Stokes source was demonstrated in the measuring of vibrational spectra of aqueous 0-10 mM glucose solutions. Scattered light was detected via photomultiplier tube and measured using either a photon counter or a lock-in amplifier in two alternative versions of the system. Both stimulated and spontaneous Raman spectra were collected with each instrument for a total of four measurement modalities. The stimulated Raman spectra measured with the photon counter showed up to 100% higher intensity for some glucose modes compared to the corresponding spontaneous Raman spectra, but also had significantly greater noise. For the spectra measured with the lock-in amplifier, the glucose modes of the stimulated Raman spectra were only 20-30% higher in intensity than those of the spontaneous Raman spectra, but had similar levels of noise. Partial least squares regression models based on spectra measured by each modality were developed and compared. The model based on stimulated Raman spectra measured with the lock-in amplifier had the strongest predictive power of all modalities and predicted the concentrations of the aqueous 0-10 mM glucose solutions with a mean squared error as low as 9.96x10-4 mM, an order of magnitude lower than that of the model based on spontaneous Raman spectra.
| Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/d8-ngfs-ry25 |
| Date | January 2019 |
| Creators | Bullen, Peter Stanley |
| Source Sets | Columbia University |
| Language | English |
| Detected Language | English |
| Type | Theses |
Page generated in 0.0027 seconds