Genetic algorithm-driven design of SERS-active surfaces for early detection of diseases

Abstract

Surface-enhanced Raman spectroscopy (SERS) enables the surface plasmon-based amplification and detection of Raman signals from biomarkers, which emerge at ultralow concentrations in the early phases of diseases. Thus, SERS chips could be used for early detection of diseases from their biomarkers obtained from liquid or tissue biopsies. While this surface enhancement capability of nanoscale gold or silver layers on different substrates were demonstrated in previous experiments and electromagnetic models, the position of the biomarker molecules on the SERS chips cannot be known or estimated a priori. As a result, SERS chips must be designed over millimeter-scale areas such that the signal amplification must be large (106 times or higher with respect to no SERS) and must span the entire slide. Simultaneous surface-enhancement of Raman signals and distributing this enhancement factor (EF) over the sample surface requires an iterative and “learning” design procedure for the geometries of nanoscale metallic features that could maximize both EF and its area simultaneously. In this study, we develop genetic algorithms and use finite-difference time-domain (FDTD) modeling to optimize the geometry of gold nanostructures (NS) on glass microscope slides to functionalize these slides as SERS-active surfaces for SERS-based enhancement of Raman spectra. By using FDTD models, we calculated the enhancement factors in 3D on glass surface for 785 nm laser for Raman spectrum measurements and used genetic algorithms (GA) to iterate on the metal NS geometry to maximize the average and the hot spot EF over the periodic patterns on the slide. Field enhancement factors as high as 1017 and 1015 were calculated for hot-spots and for whole-slide averages, respectively. The optimized structures indicate that GA could help maximize label-free and whole-slide Raman signal enhancement factors for single-cell SERS detection.