Optimizing the Parameters of Core-Shell Nanoparticles with Different Algorithms to Enhance Performance of Thin-Film Solar Cells

This computational study uses the finite-difference time-domain (FDTD) technique to investigate the use of plasmonic metal core-dielectric shell nanoparticles (NPs) of different shapes to enhance the opto-electronic performance of silicon thin-film solar cells (TFSCs). Furthermore, this study employs two machine learning-based optimization algorithms such as the Particle Swarm Optimization (PSO) algorithm and the Steepest Descent (SD) algorithm to optimize the physical parameters of the core-shell NPs to obtain maximal TFSC performance. The performances of four metal cores were investigated (aluminum, gold, silver, and titanium) and the results with parameter sweeps showed that aluminum generated the highest short-circuit current density (J
SC
) for different shapes of NPs (ellipsoid, cuboid, cylinder, and pyramid). Although, previous studies have analyzed aluminum-silica core-shell NPs on TFSCs, only limited attempts were made using machine learning-based optimizing techniques to optimize the physical parameters of the core-shell NPs. The use of a silica (SiO
2
) shell can stabilize the reactive aluminum NP core and shift its absorption spectra to longer wavelengths thus bringing favorable spectral overlap with the natural light absorption range of silicon. Overall, PSO was more effective at finding global maxima than SD. The best NP was of cuboidal shape, having an x and y span of 130.47 nm, z span of 110.12 nm, shell thickness of 12.12 nm, and FDTD x span and y span of 299.69 nm and 285.01 nm respectively. This optimal cuboidal NP brought about an improvement of 46.94% compared to a bare TFSC with no nanostructure at all.