Learning to see VOCs with Liquid Crystal Droplets

In hybrid gels with immobilized liquid crystal (LC) droplets, fast and unique optical texture variations are generated when distinct volatile organic compounds (VOCs) interact with the LC and disturb its molecular order. The optical texture variations can be observed under a polarized optical microscope or transduced into a signal representing the variations of light transmitted through the LC. We show how hybrid gels can accurately identify 11 distinct VOCs by using deep learning to analyze optical texture variations of individual droplets (0.93 average F1-score) and by using machine learning to analyze 1D optical signals from multiple droplets in hybrid gels (0.88 average F1-score).