Nanoparticles exhibit distinct optical properties due to their size, shape, and material composition, making them useful in photonics, sensing, and biological applications. Mie and Mie-Gans theory can be used to compute the optical properties of spherical and rod-shaped nanoparticles, which are limited by surface scattering and radiation damping; however, it cannot explain the optical properties of bipyramid, dumbbell, cube, and rectangular-shaped nanoparticles. Advanced computational techniques like Finite Difference Time Domain (FDTD) could be used to simulate the optical properties of complex-shaped nanoparticles, which is time-consuming and complex. To address this, we proposed to use a machine and deep learning-based forward and inverse design to explore the optical properties of gold nanorods. In forward design, we evaluated five different models and identified the XGB Regressor as the best performing model (MSE: 0.0041, MAE: 0.0190, R2: 0.994), which accurately predicts the absorption and scattering cross sections of the gold nanorod across a wavelength range of 400–1500 nm. For inverse design, we found the tandem model (MSE:0.000116, MAE: 0.004476, R2: 0.9521), which accurately identifies the length (L), width (W), and aspect ratio (AR) of the gold nanorod for achieving desired optical responses. The results provide critical insights into the underlying physics of optical effects in nanoparticles and demonstrate the capability of the integrated forward–inverse methodology for the systematic design of photonic devices and applications based on nanoparticles. The study’s findings will aid in investigating the optical properties of complex-shaped nanoparticles by leveraging the synthesis and numerical complexity and optical response. The findings will pave the way for the use of specific-size-and-shape nanoparticles with specific optical properties (such as absorption and scattering) in photothermal therapy, bioimaging, biological sensing, solar cells, photonic devices, nanoantennas, and cell biology.
Islam et al. (Wed,) studied this question.