Abstract
Molecular chirality has a key role in the helical phase structures of some liquid crystals. In particular, the induced nematic helical structure by a chiral dopant provides the macroscopic information of the dopant’s chirality at the molecular level. Developed theoretical models relate the dopants molecular chirality to their helical twisting power in a nematic host phase. Furthermore, experiments show that helical twisting power depends on temperature, solvent host, and dopant chirality. In addition, the theoretical measurements highlight a strong linear relationship between experimental helical twisting power and measured molecular chirality of some chiral dopants. We aim to create a large dataset of experimental helical twisting power measured for a wide range of nematic hosts and chiral dopants at various temperatures. The data points were collected manually using the experimental measurements. Each record consisted of the experimental molar helical twisting power, the nature of the nematic host, the nature of the chiral dopant, the molar concentration, and the temperature. The primary focus of this dataset is its reuse for machine learning predictions of the experimental helical twisting power by generating efficient chiral descriptors of dopant molecules utilised for machine learning models. In addition, the dataset aims to predict the helical twisting power of new chiral dopant molecules in various nematic liquid crystals. The primary dataset consists of 527 experimental measurements of helical twisting powers. We optimised the three-dimensional geometry of chiral dopant molecules using molecular force fields in RDKit. Then, the bootstrapping method was used to generate about 30 conformers for each primary molecule in the dataset using RDKit. The chiral descriptors were developed for use in machine-learning models for each data entry point.