Liquid crystals are responsive materials that can be used to manufacture low-cost and highly selective chemical sensors. Liquid crystals provide a potentially scalable approach toward deploying millions of wearable chemical sensors (e.g., in mobile phones or attached to clothing) that collect high-resolution data on human exposure to toxic contaminants in the air. This information is key to understanding health-risks associated with air quality, developing industrial practices that minimize workers' exposure to hazardous environments, and detecting point sources (e.g., fabrication of explosives). Liquid crystal sensors work by amplifying events that occur at the molecular-level into an optical signal when the sensor is exposed to a chemical environment. The amplification process involves a sequence of tightly coupled phenomena spanning multiple length and time scales. This span in scales lies beyond what is currently possible to characterize, model, and predict directly from first principles. This project seeks to combine first-principles and data-driven methodologies to overcome this technical challenge. The methods developed will enable the prediction of the influence of liquid crystal design variables on the information content of optical signals and will lead to a revolutionary impact on chemical sensing technologies and on the design of functional materials. The multidisciplinary nature of this project will train a new generation of engineers in the integration of data science into the design and analysis of advanced functional materials. K-12 students and the public will be engaged through development of hands-on liquid crystal sensors that respond to model target chemicals (e.g., carbon dioxide from sodas).
The project will investigate scalable machine learning techniques that enable the efficient use of large sets of experimental and first-principles simulation data to uncover and understand multi-scale phenomena that govern the performance of liquid crystals. Specifically, the project goals are to: i) Investigate the use of density functional theory and molecular dynamics simulations to identify nanoscale descriptors of the underlying spatiotemporal events occurring within and at liquid crystal interfaces (e.g., binding energies), ii) Establish feature extraction techniques to identify suitable macroscale descriptors of liquid crystal optical signals (e.g., optical response times and texture fields), and iii) Develop machine learning techniques that enable the creation of multi-scale models capable of mapping nanoscale and macroscale descriptors. These capabilities will be combined in a reinforcement learning framework that will help guide experimental data collection and identification of innovative liquid crystal system designs. The ultimate engineering goal of the project is to design LC sensors to infer exposure events involving carbon monoxide, ozone, and nitrogen and sulfur oxide.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
