Data-driven prediction and inverse design of optical asymmetry in gold nanorod-helical assemblies

The study of optical asymmetry in gold nanorod- (Au NR-) helical assemblies is of paramount importance for the development of functional nanomaterials in optoelectronics, catalysis, and biomedicine. However, such assemblies frequently proceed without adequate theoretical guidance on structure–property relationships, hindering precise control and rational design of tailored optical activity. The present study explores automated, data-driven workflows to investigate geometry-dependent optical asymmetry, with the aim of predicting the asymmetry factor (g-factor) of Au NR-helical assemblies and retrieving their geometric features. A forward artificial neural network (ANN) has been developed to predict the g-factor from geometric inputs. Conversely, a combination of ANN with particle swarm optimisation (PSO) has been demonstrated to retrieve geometric parameters necessary to achieve a target g-factor. The findings demonstrate that the forward ANN attains a high level of prediction accuracy ($r=$ 0.9833), and the inverse ANN-PSO workflow effectively identifies geometries that yield g-factors with a high degree of proximity to the target values. This demonstrates the significant value of these automated workflows for the fundamental geometric design and optical asymmetry prediction of Au NR-helical assemblies.