SOUSA, T. A. T.; http://lattes.cnpq.br/0494651199775290; SOUSA, Tiago Abreu Tavares de.
Resumo:
The objective of this study is to characterize the last layer (in which bio-chemical interactions occurs) of an optical biosensor based on the phenomenon of plasmon resonance surface in terms of its refractive index and thickness. This characterization was based on the characteristic curve of the surface plasmon resonance phenomenon and its morphological parameters, as well as an extension of the model Fresnel for polychromatic ligth sources.
The generation of the characteristic curve of the plasmon resonance phenomenon requires te treat the raw signal obtained by the iamge sensor, this work also contemplated a evaluation of digital signal processing algorithms that allows to minimize the influence of noise, maximizing quality the determination of the main attributes. the determination of the refractive index and thickness of analyte was formulated as a nonlinear optimization problem in which a computational model whose parameters were adjusted from experimental data. The neural network was designed to explore prior knowledge obtained in the characterization step, considering linear and nonlinear effects, in view of the sensitivity characteristic curve of the phenomenon of surface plasmon resonance to its morphological parameters. the model was extended to consider the polychromatic ligth, resulting in a theoretical curve closer to the experimental one, including closer morphological parameters. However this consideration implies in a loss of curve sensitivity. The techniques of signal processing applied to SPR curves must be applied in a spcific combination for greater efficiency. A neural network can estimate simultanouly the analyte refractive index and thickness, even whn trained with simulated data.