Bayesian Reconstruct of Particle Concentration and Size Distributions of Gold Nanorods from Spectral Transmittance

Gold nanorod fluids (GNRs) show promising prospects in areas such as biomedicine and energy harvesting, owing to their anisotropic morphology and tunable optical characteristics. As key parameters for understanding the physical and biological effects of nanofluids, rapid and accurate measurement of particle concentration and size within the solution is particularly crucial. Conventional characterization methods often involve complex procedures and high costs. In contrast, detection methods based on spectral inversion are limited to ideal spherical particles, rendering them unsuitable for nonspherical particles with several size parameters. To address these challenges, this study proposes a Bayesian inversion method. It establishes a computational model to derive microscopic structural parameters and the concentration of GNRs from macroscopic spectral data. By extracting characteristic information from the transmission spectrum of nanofluids, the method achieves efficient simultaneous acquisition of GNRs concentration and size distribution. The results demonstrate strong agreement between the inverted particle size values and statistical observations from transmission electron microscopy. Although the particle concentration inversion results show some discrepancies compared to measurements obtained via plasma-induced spectroscopy due to the combined effects of fluid stability and measurement errors, both remain within the same order of magnitude. Furthermore, this paper investigates the impact of discreteness in the parameter space on accuracy: the absence of particle concentration distribution constraints makes it difficult to distinguish gold nanorod particles of different sizes with similar extinction behavior, leading to unstable reconstruction results. Inversion accuracy is not sensitive to variations in short-axis size data, but is primarily constrained by the coverage range and sampling precision of aspect ratios within the data set. Coarse sampling precision and narrow sampling ranges result in significant errors in the inversion outcomes.