NSF Convergence Accelerator Track L: Intelligent Nature-inspired Olfactory Sensors Engineered to Sniff (iNOSES)

Harvard University, Cambridge MA

Investigators

Abstract

The need to acquire real-time information about the air we breathe has been brought into the spotlight through recent events and developments, including wildfires, hazardous spills, increasingly stringent emissions regulations, and the identification of specific volatiles in air, spoiled food and diseased breath. However, accurately identifying the composition of gaseous samples typically relies on bulky, expensive and stationary spectroscopic equipment. This Phase I project will introduce a portable chemical gas sensor that relies on artificial intelligence (AI) to provide highly accurate identification of volatiles in real-time. The real-time chemical sensing data will pave the way to standardization in detection and reporting across sectors – a documented challenge leading to poor accountability in emission monitoring, inefficiently timed ventilation and air purification processes, and unnecessary food waste, all of which can have atmospheric, health, and socio-economic impacts. The work will have a significant impact on STEM education, as this highly multidisciplinary project is led by a research team with a strong commitment to outreach, mentorship, and scientific communication. The concepts under investigation range from fluid dynamics, optics, and nanofabrication to AI, building simulations, and methods standardization. This creates many opportunities for research training in different disciplines, allowing students with a wide range of interests and experiences to participate. The project builds on a relatively simple sensor: a Bragg stack photonic crystal, whose optical reflection spectrum changes upon infiltration by volatiles into its pores. The time-dependence of the spectral shifts is governed by the unique transport dynamics produced by a compound or mixture of compounds and is used to continuously train a machine learning algorithm for classification and physical property prediction of compounds. Unique to this approach, the team introduces and implements olfactory inspired ‘sniffing’ sequences when volatiles are ‘inhaled’ and ‘exhaled’ in specific dynamic patterns that maximize the real-time discriminatory power of the device. Phase 1 research has included 1) prototyping, device miniaturization, and software development for detection of a subset of the target chemicals, 2) device and algorithm optimization for real-time sensing and integration into application-specific domains, guided by a combination of control theory, systems design, and machine learning, and 3) pilot studies in real environments, in partnership with industrial experts. Together, the multidisciplinary theoretical, experimental, and applied team will push the frontier of chemical sensing. 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.