Machine Learned Structure-Property Correlation Between Nanohelices and Circular Dichroism

Abstract

Rational design of chiral nanostructures with desired Circular Dichroism (CD) spectra requires a quantitative structure-property relationship, which has so far been unavailable. Using a data-driven method, the aim is to establish such a relationship for nanohelices, a prevalent structural element of chiral nanostructures. Given the challenges in synthesizing nanohelices and separating racemic mixtures, obtaining extensive CD data has been difficult. Instead, CD spectra of 1260 nanohelices are stimulated using finite-difference time-domain method. This dataset is used to train a convolutional neural network that can accurately predict the CD spectra using a few key structural parameters such as pitch and curl. Moreover, an inverse design model is developed that can generate the right helix with the desired CD. To establish quantitative relationships, Shapley Additive explanations analysis and case studies are devised for the prediction model. The algorithm efficiently analyzes the structure-property correlation, revealing the specific degrees of structural influence on the spectroscopic characteristics. Furthermore, the neural-network-based model can be extended via transfer learning to predict CD spectra of nanohelices made of other noble metals (Ag, Cu). It is believed that AI-based approaches can significantly broaden the scope of wet-chemistry nanosynthesis and computational techniques in the design of chiral nanostructures.